Back to results
US 7,925,086 B2Granted
Method and system for adaptive quantization layer reduction in image processing applications
AssigneeSamsung Electronics Co., Ltd.
InventorsZhi Zhou, Yeong-Taeg Kim
Life of this patent
Estimating the term…
Abstract
A method and a system for adaptively (dynamically) reducing quantization layer reduction for removing quantization artifacts in quantized video signals is provided. Adaptively reducing quantization layer reduction involves detecting if a selected pixel in a quantized image belongs to a ramp area in each of multiple pre-defined directions, dynamically selecting a quantization level for each of the pre-defined directions based on the corresponding detection results, and refining the pixel based on the selected quantization levels.
Claims42
A method of processing a quantized input image, comprising:
- detecting if a selected pixel in the quantized image belongs to a ramp area in each of multiple pre-defined directions by comparing luminance of pixels in the pre-defined directions to a threshold, wherein the threshold is determined based on luminance values of pixels in the pre-defined directions;
- dynamically selecting a quantization level for each of the pre-defined directions based on the corresponding detection results;
- refining the pixel based on the selected quantization levels including:
- applying a segmentation process along the ramp direction to obtain a smoothly changing area including the selected pixel;
- and
- refining the smoothly changing area based on the quantization levels to obtain a refined segment area;
- performing a low-pass filtering operation on the refined segment area to obtain a filtered area with higher precision pixel values;
- performing luminance value measurements in the refined segment area;
- and
- determining a gain value from the measured luminance values including determining a sub-gain g3 based on the neighboring pixel differences along the selected direction within the refined segment and determining the sub-gain value g3 further based on a ratio of:
- (a) the total sign changes of the neighboring differences along the pre-defined directions within the refined segment, and (b) the total number of the pixels within the refined segment.
8 more elements
The method of claim further comprising:
- comparing said ratio to threshold values, S1 and S2, wherein S1<S2;
- if the ratio is smaller than the smaller threshold S1, then setting sub-gain g3 to 1;
- and
- if the ratio is larger than the larger threshold S2, then setting sub-gain g3 to 0;
- otherwise, setting the sub-gain g3 to a linearly interpolated value between 0 and 1 based on the position of the ratio between the two thresholds S1 and S2.
2 more elements
An image processing system for processing a quantized input image, comprising:
- a hardware processor coupled with
- a detector that is configured for detecting if a selected pixel in the quantized image belongs to a ramp area in each of multiple pre-defined directions;
- a quantization level calculator that is configured for dynamically calculating a quantization level for each of the pre-defined directions based on the corresponding detection results;
- and
- a refining module that is configured for refining the pixel based on the selected quantization levels, wherein the refining module comprises:
- a pre-segmenter that is configured for applying a segmentation process along the ramp direction to obtain a smoothly changing area including the selected pixel;
- a segmenter that is configured for refining the smoothly changing area based on the quantization levels to obtain a refined segment area;
- a low-pass filter that is configured for performing a low-pass filtering operation on the refined segment area to obtain a filtered area with higher precision pixel values;
- and
- a gain controller that is configured for obtaining luminance value measurements in the refined segment area, and for determining a gain value from the measured luminance values, wherein the gain controller is further configured to determine a sub-gain g3 based on the neighboring pixel differences along the selected direction within the refined segment, wherein the gain controller is further configured to determine the sub-gain g3 based on a ratio of:
- (a) the total sign changes of the neighboring differences along the pre-defined directions within the refined segment, and (b) the total number of the pixels within the refined segment.
8 more elements
Dependencies are inferred from the claim text.