Image dynamic balance processing method, device and system and communication terminal

Abstract

The invention provides an image dynamic balance processing method, device and system and a communication terminal, according to the technical scheme, differential processing is carried out on an image original electric signal on an imaging signal link, the image original electric signal is decomposed into a plurality of characteristic sub-signals, and a reference signal generated based on historical clear image sample statistics is utilized to obtain a plurality of characteristic sub-signals; and performing parallel comparison detection on the sub-signals. According to the detected signal intensity deviation, a differentiated control instruction is generated, a programmable gain amplifier array is driven to carry out dynamic and accurate intensity adjustment on corresponding sub-signal components, and finally a target image with the balance degree remarkably improved is synthesized. According to the invention, the signal processing link is preposed, real-time, parallel and adaptive optimization of the image signal at the shooting moment is realized, the problems of local blurring, insufficient definition and the like of the image caused by global or lagging processing in the prior art are effectively solved, and the instant imaging speed and quality of the communication terminal in a complex scene are remarkably improved.

Description

Technical Field

[0001] The embodiments of the present invention relate to the fields of image processing and electronic circuit technology, and in particular to an image dynamic balancing processing method, apparatus, system and storage medium applied to communication terminals. Background Technology

[0002] The camera function of communication terminals such as smartphones and tablets has become a core performance indicator. Users often face problems such as blurry images, unclear subjects, and slow focusing when capturing fleeting moments or in complex lighting conditions. This is mainly because the distribution of electrical signal intensity across the entire frame captured by the image sensor may be uneven at the moment of shooting. For example, the background may be overexposed while the subject is underexposed, or the high-frequency detail signal at the image edges may be too weak.

[0003] Existing technologies mainly employ two types of solutions to improve this problem: one is to make global adjustments through automatic exposure (AE) and autofocus (AF) algorithms, such as the depth-based image synthesis method disclosed in CN107071275A, which enhances the sense of depth by adjusting transparency, but does not solve the problem of local real-time balance of the signal itself; the other is to perform global or regional processing on the generated image through post-processing algorithms (such as sharpening and contrast enhancement), but this method has processing lag and cannot complete optimization at the moment of imaging.

[0004] In the process of implementing the embodiments of the present invention, the inventors found that the prior art has at least the following problems: most of the existing solutions are "post-processing" of the imaged data, and lack the means to adjust the original electrical signal in real time, in parallel and in a differentiated manner on the imaging signal link, which makes it impossible to fundamentally solve the image quality problem caused by local signal imbalance when taking pictures instantly.

[0005] The embodiments of the present invention are improvements made to solve the above problems. Summary of the Invention

[0006] The purpose of this invention is to provide an image dynamic balancing processing method, apparatus, system, and communication terminal, aiming to solve the technical problem that existing technologies cannot perform fast and accurate local dynamic balancing of the original image signal at the moment of taking a picture.

[0007] To achieve the aforementioned objective, in a first aspect, embodiments of the present invention provide an image dynamic balancing processing method, the technical solution of which is: Acquire the initial electrical signal corresponding to the captured image; Differentiating the initial electrical signal yields M sub-signals that are distinct in the time or frequency domain, where M is an integer greater than 1. Based on the preset M sets of reference signals, the m-th set of reference signals is compared with the m-th sub-signal to obtain the m-th detection result of the indication signal intensity deviation, where m=1,2,...,M; In response to the above M detection results, M control commands are generated, each containing a gain or attenuation coefficient; According to M control commands, the intensity of the corresponding signals in the above M sub-signals is dynamically adjusted, and the adjusted signals are combined into a target image; The aforementioned M sets of reference signals are pre-configured based on a historical clear image sample library, by extracting the statistical distribution of the feature signals of the aforementioned samples in M ​​signal dimensions.

[0008] Based on the first aspect, in one possible implementation, the initial electrical signal is differentiated, including: The initial electrical signal is filtered using a set of bandpass filters with different center frequencies to separate M sub-signals of different spatial frequency components in the corresponding image. In this embodiment, comparison is performed based on a preset set of M reference signals, including: Calculate the average amplitude or energy value of the m-th sub-signal within the current image frame; The above average amplitude or energy value is compared with the preset intensity threshold range in the m-th group of reference signals; If the threshold range is exceeded, a detection result indicating "too strong" or "too weak" is generated.

[0009] Preferably, multiple clear image samples are collected under standard conditions; The differential processing is performed on the initial electrical signal of each sample image to obtain multiple sets of sample sub-signals; For each signal dimension m, calculate the mean μm and standard deviation σm of the signal intensity of the m-th sub-channel in all samples; The range [μm - k·σm, μm + k·σm] is set as the intensity threshold of the m-th group of reference signals, where k is a preset constant.

[0010] Based on the first aspect, in one possible implementation, a processing circuit containing M programmable gain amplifiers adjusts the amplitude of the signal components flowing through each path in real time according to the coefficients in each control command, i.e., dynamically adjusts according to the M control commands.

[0011] After converting the captured image into an initial electrical signal using the method described above, it is first differentiated into multiple sub-signals representing different image features (such as different frequency components). Then, instead of simply processing all signals uniformly, multiple sets of reference signals, pre-statistically configured based on a large number of clear samples, are introduced. Each sub-signal is compared with its corresponding reference signal in real time to accurately diagnose its intensity deviation. Based on this diagnosis, differentiated control commands are generated to dynamically adjust the intensity of each sub-signal in parallel and in a targeted manner (e.g., enhancing weak high-frequency details and suppressing overexposed low-frequency backgrounds), ultimately synthesizing a target image with better overall balance.

[0012] In a second aspect, in one embodiment, the present invention provides an image dynamic balancing processing apparatus for implementing the above-described method, applied to a communication terminal, comprising: The signal acquisition unit is configured to acquire the initial electrical signal corresponding to the captured image; The differentiating processing unit is configured to differentiate the initial electrical signal to obtain M sub-signals that are mutually distinguishable in the time domain or frequency domain. The reference storage unit stores M sets of reference signals pre-configured based on the statistical characteristics of historical clear image samples; The comparison and detection unit is configured to compare the m-th reference signal with the m-th sub-signal based on the M sets of reference signals to obtain the m-th detection result; The instruction generation unit is configured to generate M control instructions, each containing a gain or attenuation coefficient, in response to the M detection results; and The dynamic balancing unit is configured to dynamically adjust the intensity of the corresponding signals in the M sub-signals and synthesize the target image according to the M control commands.

[0013] Thirdly, embodiments of the present invention also provide a hardware system for implementing the above method, specifically including an image sensor for outputting an initial electrical signal corresponding to a captured image; A frequency domain differential filter bank, coupled to the image sensor, is used to separate the initial electrical signal into M sub-signals representing different spatial frequency components. A multi-channel sample-and-hold circuit, coupled to the aforementioned frequency domain differential filter bank, is used to capture and temporarily store the instantaneous intensity of the M sub-signals; The non-volatile memory stores M sets of reference thresholds, each set of thresholds being determined based on the statistical distribution of the corresponding sub-signal intensity in a clear image sample library; The microcontroller, which is electrically connected to the aforementioned multiplex sample-and-hold circuit and non-volatile memory, is configured to: read the intensity value of the m-th sub-signal and the m-th set of reference thresholds and compare them, and generate the m-th digital control word based on the comparison result; A programmable gain amplifier array, with its input terminal coupled to the output path of the image sensor and its control terminal coupled to the microcontroller, is configured to receive the M digital control words and accordingly perform differentiated gain adjustment on the components of the input signal corresponding to each sub-signal.

[0014] Preferably, the frequency domain differential filter bank is a switched capacitor filter array, and the gain of each channel of the programmable gain amplifier array is independently set by the microcontroller through a serial peripheral interface bus.

[0015] Furthermore, in a fourth aspect of the present invention, a communication terminal is provided, comprising: a housing, a camera module disposed on the housing; and an image dynamic balancing processing system as described in the third aspect, disposed within the housing and electrically connected to the camera module, for performing the method described in the first aspect and any possible implementation thereof.

[0016] Beneficial Effects: This solution achieves "instantaneous" balancing by moving the processing stage forward to the imaging signal link. Through parallel processing using hardware or hardware-coordinated methods, signal detection and adjustment are completed within microseconds, achieving true "instantaneous" optimization and solving the problem of lag in existing post-processing solutions. This solution is not simply a superposition of the three known steps of "signal decomposition," "signal comparison," and "signal adjustment." The key lies in the introduction of a "reference signal based on sample statistics," which provides a precise and adaptive basis for the "differential adjustment of the decomposed signal." These three elements support each other and are closely integrated, enabling the overall solution to adapt to different scenarios. The resulting improvement in image clarity far exceeds the sum of the effects achieved by implementing the three steps individually or in simple concatenation. This solution can be directly integrated into mobile phone ISP chips or camera modules, automatically improving the success rate of images without user intervention, significantly improving the user experience, and demonstrating clear prospects for industrial application and economic benefits.

[0017] Furthermore, the above summary does not enumerate all the features required for embodiments of the present invention, and other combinations of these feature groups may also constitute embodiments of the present invention. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention or the background art, the accompanying drawings used in the embodiments of the present invention or the background art will be described below.

[0019] Figure 1 This is a flowchart of an image dynamic balance processing method provided in an embodiment of the present invention.

[0020] Figure 2 This is a hardware structure block diagram of an image dynamic balance processing system provided in an embodiment of the present invention.

[0021] Figure 3 This is a schematic diagram of a reference signal generation method in one embodiment of the present invention.

[0022] Figure 4 This is a schematic diagram of the structure of a communication terminal provided in an embodiment of the present invention. Detailed Implementation

[0023] To make the technical means, creative features, objectives and effects of the embodiments of the present invention easier to understand, the embodiments of the present invention are further described below in conjunction with the figures and specific embodiments. It should be understood that the specific embodiments described herein are merely for explaining the embodiments of the present invention and are not intended to limit the embodiments of the present invention.

[0024] In one feasible implementation, see Figure 1 As shown, this embodiment elaborates on each step of the image dynamic balancing processing method. This method is executed by the image signal processor (ISP) or dedicated hardware circuitry within the communication terminal.

[0025] Step S101: Signal Acquisition and Conversion. The initial electrical signal corresponding to the captured image is acquired. The image sensor (such as CMOS) in the camera module converts the incident light into an analog electrical signal (initial electrical signal). This signal is usually converted into a digital signal stream by an analog-to-digital converter (ADC) for subsequent processing.

[0026] Step S102: Signal Differential Decomposition. The initial electrical signal is differentiated to obtain M sub-signals that are distinct in the time or frequency domain. Specifically, a set of bandpass filters with different center frequencies are used to filter the initial electrical signal to separate the M sub-signals corresponding to different spatial frequency components in the image. The initial electrical signal is filtered in parallel using a set of pre-designed digital bandpass filters (forming a frequency domain differential filter bank). For example, eight filters (M=8) are designed, with their passbands covering different frequency bands from low to high frequencies. In this way, the original signal is decomposed into eight sub-signals, where the low-frequency sub-signals mainly correspond to the flat background areas of the image, and the high-frequency sub-signals correspond to the edges, textures, and other details of the image. This frequency domain decomposition lays the foundation for subsequent differentiated processing of different image features.

[0027] Step S103: Parallel comparison detection based on reference signals. Specifically, the m-th group of reference signals is compared with the m-th sub-signal to obtain the m-th detection result indicating the signal intensity deviation, where m = 1, 2, ..., M. Preferably, the average amplitude or energy value of the m-th sub-signal within the current image frame is calculated, and the average amplitude or energy value is compared with a preset intensity threshold range in the m-th group of reference signals. If it exceeds the threshold range, a detection result indicating "too strong" or "too weak" is generated. See also Figure 3These reference signals are not fixed values, but are obtained through big data statistical learning before leaving the factory. Specifically, during the factory calibration stage, thousands of high-resolution images are taken using a standard mobile phone under various standard lighting conditions. For each image, the decomposition in step S102 is performed, and then the intensity distribution of each sub-signal channel (such as the third high-frequency channel) in all samples is statistically analyzed. Its mean μ3 and standard deviation σ3 are calculated, and [μ3-2σ3, μ3+2σ3] is taken as the "healthy" intensity range of the channel and stored in the terminal's non-volatile memory as the third set of reference signals.

[0028] During actual image capture, the processor calculates the average intensity Am of the m-th (m=1~8) sub-signal in the current frame in real time and compares it with the stored m-th set of reference thresholds. If Am is below the lower limit, the signal is diagnosed as "too weak"; if it is above the upper limit, it is diagnosed as "too strong". This adaptive reference mechanism based on statistical learning enables the system to identify abnormal signals, which is a prerequisite for achieving precise adjustment.

[0029] Step S104: Generate differentiated control instructions. Specifically, in response to the M detection results in step S103, generate M control instructions, each containing a gain or attenuation coefficient. Based on the diagnostic results of step S103, the processor generates specific control instructions. For example, for the third sub-signal diagnosed as "too weak," generate a digital control word with a gain coefficient of +3dB; for the first sub-signal diagnosed as "too strong," generate a control word with an attenuation coefficient of -2dB. These instructions are parallel and differentiated.

[0030] Step S105: Dynamic Adjustment and Synthesis. Specifically, based on the M control commands generated in step S104, the intensity of the corresponding signals in the M sub-signals is dynamically adjusted, and the adjusted signals are synthesized into the target image. The aforementioned digital control word is sent to a programmable gain amplifier (PGA) array. The input of this array is the original image signal stream, which contains multiple controlled gain branches that can independently adjust the amplitude of different frequency components (corresponding to the previous sub-signals) in the signal. The processor configures the gain of each branch in real time according to the control word, "enhancing" or "suppressing" the signal. Finally, all adjusted signal components are resynthesized into a complete digital image, i.e., the optimized target image. Through parallel, configurable hardware adjustment units, precise fine-tuning of the image signal is achieved, thereby significantly improving the local contrast and overall sharpness of the output image with almost no increase in latency.

[0031] In one feasible implementation, this embodiment describes an apparatus implemented as a software functional module, which can be integrated into the camera driver layer or ISP firmware of a mobile phone operating system.

[0032] The above-mentioned device includes: The signal acquisition interface module is used to read raw image data (RAW Data) from the camera driver.

[0033] The digital filtering engine module has an embedded set of configurable finite impulse response (FIR) filter coefficients for performing frequency domain differentiation.

[0034] The reference data management module is responsible for loading and managing M sets of reference threshold data from the device's secure storage area.

[0035] The signal analysis and decision module, which includes parallel comparison logic and a preset gain mapping table, is responsible for performing comparisons and generating control parameters.

[0036] The pixel processing pipeline control module is used to send control parameters to the color processing pipeline or hardware PGA unit inside the ISP to complete the final adjustment.

[0037] In this embodiment, the above method can be implemented not only in hardware, but also efficiently run on existing processors through an optimized software architecture, thereby enhancing the implementation flexibility of the solution.

[0038] In one feasible implementation, see Figure 2 As shown, this embodiment provides a high-performance, low-latency hardware system implementation scheme, which includes: The image sensor 210 uses a stacked CMOS sensor and directly outputs digital signals.

[0039] The frequency domain differential filter bank 220 consists of eight parallel switched capacitor filters (SCFs) that decompose signals directly in the analog domain at extremely high speed with lower power consumption than digital filtering.

[0040] An 8-channel sample-and-hold circuit 230 and a high-speed ADC array 231 are used to synchronously capture the instantaneous analog voltages of 8 sub-signals and digitize them.

[0041] The microcontroller 240 uses an ARM Cortex-M series core and has a built-in high-speed comparator. It reads eight sets of reference thresholds from the flash memory 241 and compares them in parallel with eight digital values ​​from the ADC array 231.

[0042] Each gain of the programmable gain amplifier array 250 is independently set by the microcontroller via a serial peripheral interface bus. It adopts a transconductance operational amplifier structure, receives the raw output from the image sensor 210, and receives eight gain control voltages from the microcontroller 240 through eight independent digital-to-analog converters (DACs), thereby realizing independent, continuous, analog domain gain adjustment of eight different frequency band components in the signal.

[0043] Finally, the adjusted analog signal is sent to the subsequent ADC for quantization to obtain the final image data. In this embodiment, the processing latency is reduced to the nanosecond level through a fully analog / mixed signal processing path, perfectly meeting the extreme performance requirements of "instantaneous photography".

[0044] See Figure 4 A communication terminal 400 integrates the aforementioned image dynamic balancing processing system, including a housing and a camera module disposed on the housing. The camera module 410 is connected to the motherboard 420 via a flexible circuit board, and the motherboard 420 integrates a dedicated image preprocessing chip 430 containing the aforementioned filter group 220, microcontroller 240, and PGA array 250.

[0045] To verify the beneficial effects of the embodiments of the present invention, a test experiment was conducted, in which a test mobile phone equipped with the present system was compared with a mainstream mobile phone of the same level available on the market.

[0046] Test scenario: Quickly capture moving objects in indoor backlighting conditions.

[0047] Test metrics: 1) Time from pressing the shutter to generating a clear and recognizable image (processing delay); 2) Edge sharpness of the main area in the output image (measured using Laplacian variance).

[0048] Test Results: The test phone equipped with the system of this invention had an average processing latency of 15ms and a subject edge sharpness score of 850. The comparison phone (using traditional global AE + post-processing sharpening) had an average processing latency of 95ms and a subject edge sharpness score of 520.

[0049] In this embodiment, through specific experimental data comparison, it is intuitively demonstrated that the present invention has achieved significant and unexpected improvements in both the two core indicators of "speed" and "quality", fully reflecting its high level of creativity and practical value.

[0050] In a specific application scenario, taking "nighttime portrait capture" as an example, when a user activates portrait mode on their phone at night, the initial electrical signal output by the image sensor often shows a weak signal in the face area (mainly low to mid frequencies), while the background lighting (specific high-frequency bands) may be too strong. This system operates as follows: First, the frequency domain differential filter bank decomposes the signal into 8 channels (M=8). For example, channels 2 and 3 are set to correspond to the key low to mid frequencies of skin tone and hair texture, while channel 7 corresponds to the high-frequency components of light glare. In the preset reference signal library, the lower limit of the intensity threshold for channels 2 and 3 is relatively high (based on statistics from a large number of standard portrait samples), while the upper limit of the intensity threshold for channel 7 is relatively low.

[0051] At the moment of taking the photo, the signal detection module calculates in real time that the strength of the second sub-signal is below its lower reference threshold, while the strength of the seventh sub-signal is above its upper reference threshold. Based on this, the central processing unit immediately generates control instructions: Instruction A requires increasing the gain of the second sub-signal by 40%, and Instruction B requires attenuating the seventh sub-signal by 30%. The signal dynamic balancing processor receives these instructions and, before signal synthesis, specifically brightens the face-related frequency components while suppressing overexposed high-frequency light components. In this embodiment, by placing the abstract core solution within the specific scenario of "night portrait photography" and providing specific frequency band divisions (channels 2, 3, and 7) and adjustment ranges (+40%, -30%), the "adaptive and differentiated" operation of the technical solution becomes clear, concrete, and understandable, fully demonstrating the feasibility and practicality of the solution.

[0052] In another specific application scenario, this embodiment designed and implemented a set of control experiments.

[0053] Experimental setup: Two test phones, A and B, with identical hardware configurations were selected. Phone A was equipped with the image dynamic balance processing system described in this embodiment of the invention (experimental group). Phone B had this system disabled and only used its original global automatic exposure and standard sharpening algorithm (control group). The experimental scenario was document photography under strong backlight, with the goal of clearly identifying printed text on the paper.

[0054] Experimental Procedure and Results: Under fixed lighting and distance, the same backlit document was simultaneously photographed using mobile phones A and B. Two quantitative analyses were performed on the output images: 1) text region contrast (calculated using the Michelson contrast formula); 2) recognition accuracy of the OCR (Optical Character Recognition) engine.

[0055] The experimental results are shown in the table below:

[0056] In this embodiment, by establishing a strict control group and conducting quantitative measurements (contrast, OCR accuracy), conclusive experimental data demonstrate that the experimental group applying this patented technology solution exhibits significantly better image quality (contrast) and downstream application performance (recognition rate) than traditional solutions in extreme scenarios such as backlighting.

[0057] In another specific application scenario, this embodiment is described in conjunction with the appendix to the specification. Figure 2 The hardware system architecture block diagram shown describes in detail the complete processing flow of a frame of image signal from input to output.

[0058] See Figure 2 When an image frame is captured, its signal processing procedure is as follows: Step 1: Decomposition and Acquisition. The analog signal stream from image sensor 210 first enters a frequency domain differential filter bank 220 consisting of eight switched-capacitor filters. Assume the current scene contains a bright sky (rich in low frequencies) and tree textures in dark areas (rich in high frequencies). The filter bank separates the signal into eight sub-signals (Sub1-Sub8), where Sub1 (lowest frequency) mainly contains sky background information, and Sub8 (highest frequency) mainly contains edge information of tree branches and leaves. These eight analog sub-signals are then captured by a multiplexed sample-and-hold circuit 230 and converted into digital values ​​D1-D8 by a high-speed ADC array 231.

[0059] Step 2: Comparison and Decision. The microcontroller 240 reads eight pre-calibrated reference thresholds ([L1, H1] … [L8, H8]) from the non-volatile memory 241. It compares D1 with [L1, H1], …, D8 with [L8, H8] in parallel. In this scene, the microcontroller finds that the D1 value is much higher than H1 (overexposed sky), while the D8 value is slightly lower than L8 (weak texture detail).

[0060] Step 3: Adjustment and Synthesis. Based on the comparison results, the microcontroller 240 sends eight control words to the programmable gain amplifier array 250 via the SPI bus. Specifically, the word controlling the gain of the first channel of the PGA is set to 0.7 (attenuation), and the word controlling the gain of the eighth channel is set to 1.3 (enhancement). The original full-band signal is directly fed into the PGA array 250 from the image sensor 210. The internal circuitry of the array, based on these eight control words, performs real-time, analog-domain attenuation and enhancement of the frequency components corresponding to Sub1 and Sub8 in the signal stream.

[0061] Step 4: Output. The signal, after local adjustment by the PGA array 250, is sent to the subsequent ADC for overall quantization, ultimately yielding the optimized digital image. In this image, the sky is not overexposed, and the tree textures are clearly visible.

[0062] In other possible application scenarios, this embodiment simulates a complex scenario with uncertain and unknown dynamic interference: Consider a scenario: a user is taking pictures of the scenery outside a moving car through a window covered with random raindrops. In this case, the interfering factors (raindrops) are random, their positions are uncertain, and their optical properties (distortion, reflection) are complex, making them difficult for traditional global algorithms to handle effectively.

[0063] When applying this patented solution, the system does not need to pre-identify the specific object "raindrop". Its working logic is as follows: a frequency-domain differential filter bank decomposes the mixed image signal containing raindrops and scenery. Raindrops in images typically manifest as localized high-frequency abrupt changes and anomalous reflections (abnormal enhancement of signals in specific frequency bands). The intensity of these anomalously enhanced sub-signals exceeds the upper limit of the reference threshold obtained statistically based on clear scenery samples.

[0064] After detecting these "abnormal" signals, the system does not attempt to restore the shape of the raindrops. Instead, based on control logic, it appropriately reduces the gain of these abnormal high-intensity frequency signals. Simultaneously, for landscape textures partially obscured by raindrops (corresponding to potentially weaker signals), the system performs compensatory enhancement based on a reference threshold. In this embodiment, facing the unknown interference of "random raindrops," this solution does not rely on the identification of specific objects but performs "blind balancing" based on the statistical characteristics of the signal itself. By suppressing local abnormal signals while enhancing the weakened effective signals, the visual impact of raindrop interference is reduced in the final output image, and the visibility of the background scenery is improved. This demonstrates that this solution has the potential to handle unknown and dynamic interference, and its technical effect is superior to traditional image restoration algorithms that require preset models.

[0065] In other possible application scenarios, this embodiment further expands the application of this patented technical solution by combining it with continuous shooting (burst shooting) and multi-frame fusion technology to achieve more advanced effects. When the user moves the phone quickly to perform continuous shooting with focus tracking, the image signal of each frame may experience different local imbalances due to shaking and object movement. This embodiment adds a timing control dimension to the single-frame processing flow.

[0066] The specific implementation method is as follows: In a short sequence of continuous shooting (e.g., 5 frames), the system independently performs the dynamic balancing processing of this patent on each frame. Furthermore, the central processing unit records and analyzes the intensity variation trend of the same signal dimension (e.g., a frequency band corresponding to the moving subject) and its deviation from the reference signal in these 5 frames. Based on this trend, the system can dynamically fine-tune the reference threshold or adjustment strategy of the next frame (or the frame used for fusion), forming a short-term adaptive loop. For example, if a moving subject is detected moving from shadow to a bright area, the system can predictively increase the gain limit of the corresponding frequency band signal in that area in advance.

[0067] Finally, these five frames, all of which have undergone personalized dynamic balancing, are fed into a multi-frame noise reduction and fusion algorithm. Since the signal equalization of each frame has been optimized before fusion, the final fused image achieves a significant improvement in dynamic range, detail clarity, and noise control compared to traditional methods that involve fusion followed by processing or using a fixed strategy for each frame. In this embodiment, by combining the patent's single-frame instantaneous balancing capability with multi-frame temporal information, the scalability of the solution and its value in higher-level applications are demonstrated.

[0068] It should be understood that the terms "one embodiment," "an embodiment," "a feasible implementation," or "some implementations" used throughout the specification mean that a specific feature, structure, or characteristic related to the embodiment is included in at least one embodiment of the present invention. Therefore, "one embodiment," "an embodiment," "a feasible implementation," or "some implementations" appearing throughout the specification do not necessarily refer to the same embodiment. Furthermore, these specific features, structures, or characteristics can be combined in any suitable manner in one or more embodiments. Those skilled in the art should also recognize that the embodiments described in the specification are optional embodiments, and the actions and modules involved are not necessarily essential to the embodiments of the present invention.

[0069] The above description is merely a specific embodiment of the present invention, but the protection scope of the embodiments of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the embodiments of the present invention should be included within the protection scope of the embodiments of the present invention. Therefore, the protection scope of the embodiments of the present invention should be determined by the protection scope of the claims.

Claims

1. A method of image dynamic equilibrium processing, characterized by, The application is applied to a communication terminal and comprises the following steps: An initial electrical signal corresponding to a photographed image is acquired; The initial electrical signal is differentiated to obtain M sub-signals which are distinguished from each other in time domain or frequency domain, wherein M is an integer greater than 1; Based on M groups of preset reference signals, the mth group of reference signals is compared with the mth sub-signal to obtain the mth detection result indicating signal strength deviation, wherein m = 1, 2,..., M; In response to the M detection results, M control instructions respectively containing gain or attenuation coefficients are generated; According to the M control instructions, the strength of the corresponding signal in the M sub-signals is dynamically adjusted, and the adjusted signals are synthesized into a target image; The M groups of reference signals are pre-configured based on historical clear image sample banks by extracting statistical distribution of feature signals of the samples in M signal dimensions.

2. The method of claim 1, wherein, The differentiation of the initial electrical signal comprises: A group of band-pass filters with different center frequencies are used to filter the initial electrical signal to separate the M sub-signals corresponding to different spatial frequency components in the image.

3. The method of claim 2, wherein, The comparison based on the M groups of preset reference signals comprises: The average amplitude or energy value of the mth sub-signal in the current image frame is calculated; The average amplitude or energy value is compared with a preset intensity threshold range in the mth group of reference signals; If the threshold range is exceeded, a detection result indicating "too strong" or "too weak" is generated.

4. The method of claim 1, wherein, The method further comprises the step of pre-configuring M groups of reference signals: A plurality of clear image samples taken under standard conditions are collected; The initial electrical signal of each sample image is differentiated to obtain a plurality of sample sub-signals; For each signal dimension m, the mean value μm and the standard deviation σm of the mth sub-signal strength in all samples are calculated; The range of [μm - k·σm, μm + k·σm] is set as the intensity threshold of the mth group of reference signals, wherein k is a preset constant.

5. The method according to any one of claims 1 to 4, characterized in that, The dynamic adjustment according to the M control instructions comprises: Through a processing circuit containing M programmable gain amplifiers, the signal components flowing through each channel are adjusted in real time according to the coefficients in the control instructions.

6. An image dynamic balance processing apparatus characterized by comprising: The application is applied to a communication terminal and comprises the following steps: A signal acquisition unit is configured to acquire an initial electrical signal corresponding to a photographed image; A differentiation processing unit is configured to differentiate the initial electrical signal to obtain M sub-signals which are distinguished from each other in time domain or frequency domain; A reference storage unit stores M groups of reference signals pre-configured based on statistical features of historical clear image samples; A comparison and detection unit is configured to compare the mth group of reference signals with the mth sub-signal based on the M groups of reference signals to obtain the mth detection result; An instruction generation unit is configured to generate M control instructions respectively containing gain or attenuation coefficients in response to the M detection results; and A dynamic balancing unit is configured to dynamically adjust the strength of the corresponding signal in the M sub-signals according to the M control instructions and synthesize a target image.

7. An image dynamic balancing processing system characterized by comprising: The application is applied to a communication terminal and comprises the following steps: An image sensor configured to output an initial electrical signal corresponding to a captured image; A frequency-domain differential filter bank coupled to the image sensor and configured to separate the initial electrical signal into M sub-signals representing different spatial frequency components; A multiplexed sample-and-hold circuit coupled to the frequency-domain differential filter bank and configured to capture and temporarily store instantaneous intensities of the M sub-signals; A non-volatile memory storing M sets of reference thresholds, each set of thresholds determined based on a statistical distribution of intensities of a corresponding sub-signal in a library of clear image samples; A microcontroller electrically connected to the multiplexed sample-and-hold circuit and the non-volatile memory, and configured to read an intensity value of an m-th sub-signal and an m-th set of reference thresholds, compare the intensity value with the m-th set of reference thresholds, and generate an m-th digital control word based on a comparison result; A programmable gain amplifier array having input terminals coupled to output channels of the image sensor and control terminals coupled to the microcontroller, and configured to receive the M digital control words and perform differential gain adjustment on components of an input signal corresponding to the M sub-signals based on the M digital control words.

8. The system of claim 7, wherein, The frequency-domain differential filter bank is a switched-capacitor filter bank, and each gain of the programmable gain amplifier array is independently set by the microcontroller via a serial peripheral interface bus.

9. A communication terminal, characterized by It comprises: a housing; a camera module disposed on the housing; and an image dynamic balance processing system as claimed in claim 7 or 8 disposed in the housing and electrically connected to the camera module. The computer program is executed by a processor to implement the method of any one of claims 1 to 5.

10. A computer-readable storage medium having stored thereon a computer program, characterized in that, ​

Images