Video Preprocessing Temporal Spatial Filter for Noise Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Digital video compression systems face inefficiencies due to random noise, such as temporal and salt-and-pepper noise, which are indistinguishable from high-frequency image details, leading to degradation in frame quality and wastage of bits for encoding noise rather than actual information.
Innovation Solution
A method involving a filter that calculates an activity value for each pixel based on the sum of squared differences between adjacent pixels in current and previous frames, with pixel values modified if the activity does not exceed a threshold, using an alpha blending technique to reduce noise while preserving actual high-frequency information.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If a low-pass filter is used to reduce high-frequency noise, then noise reduction is improved, but actual high-frequency image details are lost resulting in a soft image
Solution Approach 1:
The filter dynamically adapts its behavior based on local activity measurement. For each pixel, the filter calculates an activity value representing local high-frequency content, and adjusts the filtering strength accordingly - applying strong filtering only in low-activity regions while preserving details in high-activity regions
Solution Approach 2:
The filter applies different processing to different regions of the image based on local characteristics. Each pixel is processed independently with filtering strength determined by its local activity value, allowing noise reduction in smooth areas while preserving edges and details in active areas
2Object-affected harmful factors
If a two-dimensional spatial filter is used to reduce noise, then noise reduction is improved, but detection of borders and edges is affected by noise
Solution Approach 1:
The filter dynamically adjusts its operation based on local activity values calculated for each pixel. High-activity regions (edges/borders) are identified and preserved, while low-activity regions (smooth areas) undergo noise reduction
Solution Approach 2:
Different filtering operations are applied to different pixels based on their local activity. Edge pixels with high activity values are preserved without filtering, while smooth region pixels with low activity values receive noise reduction
3Loss of information
If all high-frequency content is processed by the compression system, then complete information encoding is achieved, but bits are wasted encoding noise instead of actual information
Solution Approach 1:
The filter extracts and removes noise components from the image before compression, separating actual image information from noise. This allows the compression system to encode only meaningful information, improving efficiency
Solution Approach 2:
Noise reduction is performed as a preliminary step before compression encoding. By removing noise beforehand, the subsequent compression process operates on cleaner data, reducing the bit stream size while maintaining quality
Data Source
AI summary
The invention refers to an apparatus and a method for reducing random noise in a sequence of digital video frames comprising the following steps:1. for each of the pixels (center pixel) in a frame a set of adjacent pixels is defined;2. for each of the adjacent pixels the difference of their values in the current frame and the previous frame is calculated, whereby the value of the center pixel is omitted;3. each difference value is shifted right for a predefined number of bits;4. the square of the difference value is added to an activity value of that center pixel;5. if the activity value remains below a predefined threshold value, then a weighting factor depending from activity value is calculated and6. the value of the center pixel is set to a weighted value.