30 results about "Nonlinear normalization" patented technology

The present invention discloses a quality estimation method of a parametric image based on nonlinear structural similarity deviation. Firstly the RGB color image spaces of a reference image and a degraded image are converted into a Gauss image space and a grayscale image space, then a local edge intensity spectrum and a local gradient spectrum are generated and are subjected to nonlinear normalization, the corresponding local edge similarity map and the corresponding local gradient similarity map are calculated, and finally through analyzing the structural characteristic of the local similarity map, a value with a small similarity deviation is adaptively selected to be the quality estimation value of the degraded image. According to the method, the quality estimation effect of different fuzzy, JPEG, noise and other natural images is good, the calculation is convenient and efficient, and the realizability is good.

The invention relates to an iris texture normalization method based on a dual-spring model. The method includes that ration of pupil radius to iris radius is set under an iris standard state, and ratio of the width of a near-pupil to the width of (near-pupil area + far-pupil area) is set under an iris standard state; the number of required sampling lines of the near-pupil area and the far-pupil area is set; the near-pupil area and the far-pupil area are modeled into series connected springs with different elastic coefficients; lower-pupil radius, near-pupil area width and far-pupil area widthunder a standard state are calculated according to iris actual radius; inner circumference and outer circumference of an iris are divided equally, correspondingly and respectively, coordinates of a certain point on the edge of a pupil and coordinates of a point arranged on the edge of the iris and corresponding to the point on the edge of the pupil are calculated, and distance between the two points is calculated; when the iris is under the non-standard state, the width of the near-pupil area and the far-pupil area is calculated, and linear sampling is conducted on the near-pupil area and thefar-pupil area; and sampling points are normalized into a rectangular image. Compared with the existing non-linear normalization method, the method is simple and practical, can reduce errors caused by pupil contracting and expanding, and can increase recognition rate of iris recognition.

The invention discloses a background modeling method based on pixel confidence, which comprises judging whether a pixel point is a background according to the pixel confidence, that is, the possibility that the pixel point is the background, specifically comprising the following steps of: firstly reading video frames in order, judging the size of an HSV pixel value at a corresponding point of twoadjacent frame images so as to judge whether the point is stable, and if so, adding a continuous stable count value; otherwise, zeroing the continuous stable count value at the pixel point; then performing the non-linear normalization process on the continuous stable count values of all pixel points on the image to obtain unmodified confidence at each pixel point on the current image; then clustering background candidate list colors at each pixel point in the image and updating the cluster confidence; and finally taking a pixel value with the highest cluster confidence as a current background, continuously reading the next frame image. The background can be accurately extracted to get ready for obtaining a monitored object; and the background modeling method can be widely applied to the video monitoring fields such as object tracking, movement analysis and the like.

The present invention discloses a quality estimation method of a parametric image based on nonlinear structural similarity deviation. Firstly the RGB color image spaces of a reference image and a degraded image are converted into a Gauss image space and a grayscale image space, then a local edge intensity spectrum and a local gradient spectrum are generated and are subjected to nonlinear normalization, the corresponding local edge similarity map and the corresponding local gradient similarity map are calculated, and finally through analyzing the structural characteristic of the local similarity map, a value with a small similarity deviation is adaptively selected to be the quality estimation value of the degraded image. According to the method, the quality estimation effect of different fuzzy, JPEG, noise and other natural images is good, the calculation is convenient and efficient, and the realizability is good.

The invention discloses a nonlinear normalization based IQA (image quality assessment) method of a Laplace-Gaussian signal. The method comprises steps as follows: redundancy elimination expression is performed on images firstly and is completed through two processes including LOG filtering and nonlinear normalization; the two processes are used for eliminating first-order and second-order statistical redundancy in the images and reducing high-order statistical redundancy respectively; two computing methods are proposed to predict the subjective quality or the distortion degree of the images, and the two computing methods are marked as NLOG-MSE (mean square error) and NLOG-COR (correlation) respectively; according to NLOG-MSE, mean square errors between the original images subjected to redundancy elimination expression and test images is computed to obtain the distortion measurement of the test images, and according to NLOG-COR, correlation between the original image and test image in each point and redundancy elimination expression is computed to predict the image quality. The experimental result proves that the two computing methods have good prediction performance in the aspect of IQA, the NLOG-MSE method has simple computation, and the application of the NLOG-MSE method in other fields is greatly facilitated.

The invention discloses a method for automatically extracting interesting image regions based on a human visual attention system, which mainly solves the problem that the prior method for extracting interesting regions cannot extract a plurality of interesting regions and edge information. The method comprises the following steps: calculating the local brightness contrast, global brightness contrast and edge of an input image; fusing characteristic graphs corresponding to the three characteristics through global nonlinear normalization merging algorithm so as to generate a contrast graph; calculating the position characteristics of the input image so as to establish a weighting graph; establishing a saliency graph corresponding to the input image through the contrast graph and the weighting graph; and segmenting out the interesting regions of the input image according to the saliency graph. The method can effectively extract a plurality of interesting regions in the input image, and is used in the technical field of image analysis and image compression.

The invention relates to a method for pre-processing an iris image. The method comprises the following steps of: (1) extracting an infrared lamp image in the iris image; (2) calculating the size and the centre position of a reflection imaging light aperture; (3) carrying out low gray level filling on an illumination light reflection image region; (4) selecting an image in a certain range at four sides as a local image including pupils by taking the centre of the aperture as the centre; (5) precisely locating an inner boundary; (6) precisely locating the boundary of an iris; and (7) nonlinearly normalizing an iris region: firstly, calculating the expansion rate of the iris, determining a quantized interval by using the expansion rate, and then normalizing irregular annular regions into regular rectangular regions. The method for pre-processing the iris image disclosed by the invention cannot be influenced by light sources, light spots and shading of eyelids and eyelashes so as to avoid wrong orientation. The method has strong algorithm robustness and high calculation speed. The method can be used for processing the normalization process of the iris image in real time.

The invention discloses a method for defogging based on double-tree complex wavelet, which belongs to the technical field of image processing. In the first step of the present invention, the approximate value of atmospheric ambient light is obtained through dual-tree complex wavelet transform, and then the second step uses the improved dark channel to derive the nonlinear normalized atmospheric transmittance map, and then estimates the fog map, and finally according to the original map , to eliminate the fog and form a fog-free map. The invention has better effect on removing natural images such as uniform fog.

The invention discloses a vegetation index re-synthesis algorithm, and belongs to the technical field of vegetation coverage change research. According to the method, EVI and NDVI are resynthesized interms of time and space, and the method specifically comprises the following steps: extracting monthly NDVI and monthly EVI of a to-be-detected region within nearly 5-10 years; synthesizing a monthlyNDVI value and an EVI value by using a maximum value synthesis method and a mean value synthesis method, and determining an EVIMax value, an EVIMean value, an NDVIMax value and an NDVIMean value as basic data of the synthesized VI through correlation analysis; carrying out nonlinear normalization processing on the EVIMax value, the EVIMean value, the NDVIMax value and the NDVIMean data, and calculating a synthetic vegetation index VI. The vegetation cover change analyzed by the synthetic vegetation index has better research effect than a single vegetation index.

The invention discloses a method and a device for establishing a numerical fusion model. The method comprises: obtaining numerical fusion dimensionality information; according to data distribution characteristic of each piece of dimensionality information of numerical fusion, using a normalization algorithm corresponding to each piece of dimensionality information, calculating to obtain a normalization numerical value corresponding to each piece of dimensionality information; analyzing logical relationship of each piece of dimensionality information, performing dimensionality fusion calculation on two normalization numerical values of dimensionality information, the normalization numerical values satisfying a preset logical relationship condition, to obtain a dimensionality fusion calculated value; performing multiple-dimension numerical fusion processing on the dimensionality fusion calculated value and the normalization numerical value of residual dimensionality information, to establish and generate a numerical fusion model. The method and the device realize objectives of nonlinear normalization processing and multiple-dimension comprehensive accurate evaluation.

The invention discloses a method for recognizing an SAR target based on curvelet transform, and relates to the technical field of image processing and target recognition. The invention mainly solves the problems of low recognition rate and large volume of data existing in the prior method for recognizing the SAR target. The implementation process comprises the followings steps of: smoothing medianfiltering; performing nonlinear normalization; performing threshold segmentation after window filtering; extracting low-frequency subband information and the threshold segmentation by the curvelet transform; forming a fine characteristic area by morphological treatment; orderly performing the smooth filtering, the nonlinear normalization, the window filtering and the extraction of the low-frequency subband information of an image by the curvelet transform to form a rough characteristic area; combining the fine characteristic and the rough characteristic as a final target characteristic vectorto train and support a vector classifier; and recognizing the SAR target by the well trained classifier. The method has the advantages of reducing the volume of data and improving the recognition rate, and can be applied to the identification of SAR ground targets.

The invention relates to an iris texture normalization method based on a dual-spring model. The method includes that ration of pupil radius to iris radius is set under an iris standard state, and ratio of the width of a near-pupil to the width of (near-pupil area + far-pupil area) is set under an iris standard state; the number of required sampling lines of the near-pupil area and the far-pupil area is set; the near-pupil area and the far-pupil area are modeled into series connected springs with different elastic coefficients; lower-pupil radius, near-pupil area width and far-pupil area width under a standard state are calculated according to iris actual radius; inner circumference and outer circumference of an iris are divided equally, correspondingly and respectively, coordinates of a certain point on the edge of a pupil and coordinates of a point arranged on the edge of the iris and corresponding to the point on the edge of the pupil are calculated, and distance between the two points is calculated; when the iris is under the non-standard state, the width of the near-pupil area and the far-pupil area is calculated, and linear sampling is conducted on the near-pupil area and the far-pupil area; and sampling points are normalized into a rectangular image. Compared with the existing non-linear normalization method, the method is simple and practical, can reduce errors caused by pupil contracting and expanding, and can increase recognition rate of iris recognition.

The invention relates to a stereoscopic video saliency detection method based on a binocular multidimensional perception characteristic. A traditional model method cannot effectively detect a salient region of a stereoscopic video. The method comprises the steps of salient feature extraction and salient feature fusion; the step of salient feature extraction is as follows: respectively performing saliency calculation from view information of the stereoscopic video of three different dimensions of space, depth and motion, including two dimensional static salient region detection, depth salient region detection, and motion salient region detection; and the step of salient feature fusion is as follows: using a global nonlinear normalized fusion strategy to perform fusion on an acquired salient feature map of three different dimension, so as to acquire a stereoscopic video salient region. According to the stereoscopic video saliency detection method based on the binocular multidimensional perception characteristic provided by the invention, the computational complexity is low, the quality of an acquired stereoscopic video saliency map is high, and the method can be directly applied to 3D video compression, 3D quality assessment and object identification and tracking and other engineering fields.

The invention discloses a new energy power generation power prediction method and system with an adaptive time scale, including: step 1: perform power prediction at equal intervals according to external meteorological data; step 2: calculate adjacent power prediction results according to step 1 The absolute value of the power slope between two power prediction time points; Step 3: Perform nonlinear normalization processing on the absolute value of the slope; Step 4: Calculate the prediction time interval according to the results of the nonlinear normalization processing; Step 5: According to the step 4. Calculate the generated power data of each prediction point separately for the obtained prediction time interval. In the present invention, when the change trend of wind power and photovoltaic power generation is relatively large, the frequency of power prediction is increased; when the change trend of wind power and photovoltaic power generation is stable, the frequency of power prediction remains unchanged, which solves the problem that the existing prediction system has a fixed prediction time interval. Accurately reflect the changing trend of wind power and photovoltaic power generation.