Image Coding Orthogonal Transformation Size Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional image coding methods face challenges in accurately detecting character blocks, leading to decreased coding efficiency and incorrect selection of orthogonal transformation sizes, particularly in natural images and images with edges.
Innovation Solution
An image coding method that calculates histograms and detects plane regions within blocks to determine the appropriate orthogonal transformation size, considering both bimodal distribution and plane region size, and uses neighbor block information to adjust threshold values for accurate character detection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If edge detection based on adjacent pixel difference is used to select orthogonal transformation size, then character portion detection is attempted, but detection accuracy is low and incorrect selection occurs in natural images
Solution Approach 1:
The invention changes the detection parameter from adjacent pixel difference to luminance histogram bimodal distribution. By analyzing whether the luminance histogram shows bimodal distribution (having two peaks), the system can accurately detect character blocks. This parameter change resolves the contradiction by providing both high detection accuracy and maintained coding efficiency, as the histogram method correctly identifies character portions without mistakenly selecting blocks in natural images.
2Object-affected harmful factors
If orthogonal transformation size of 4 pixels×4 pixels is selected for blocks with edges, then mosquito noise is suppressed, but frequency resolution decreases and coding efficiency drops in images with high correlation
Solution Approach 1:
The invention uses luminance histogram bimodal distribution as the selection criterion instead of edge detection. This allows accurate identification of character blocks where mosquito noise occurs, enabling selection of 4 pixels×4 pixels transformation size only where needed. The method maintains high frequency resolution in natural image regions by using 8 pixels×8 pixels transformation, thus suppressing mosquito noise without sacrificing coding efficiency.
3Measurement precision
If threshold value for edge detection is set low to detect all character blocks, then detection coverage increases, but incorrect selection of blocks with leaves occurs
Solution Approach 1:
The invention changes from using edge detection with adjustable threshold to using luminance histogram bimodal distribution analysis. This parameter change inherently provides reliable detection without false positives. The bimodal distribution characteristic is specific to character blocks and does not occur in natural image blocks with leaves, ensuring both accurate detection coverage and high reliability without needing to adjust threshold values.
Data Source
AI summary
The image coding method according to the present invention includes the following steps. At a histogram calculation step, a histogram of pixel values included in a block image is calculated. At a plane region detection step, a size of a plane region included in the block image is detected. At an orthogonal transformation size selection step, (i) a first processing size is selected as a processing size of orthogonal transformation for the block image when the histogram is bimodal and the size of the plane region is equal to or greater than the first threshold value, and (ii) a second processing size greater than the first processing size is selected as the processing size, when that the histogram is not bimodal or the size of the plane region is smaller than the first threshold value.