Low complexity perceptual visual quality evaluation for JPEG2000 compressed streams
A method of detecting image quality in a wavelet transform-encoded image includes receiving wavelet transform-encoded image data partitioned into tiles, each tile partitioned into a number of subbands, evaluating a number of wavelets in fewer than the number of subbands, assigning a measure to each of the subbands evaluated for each tile, using the measures to determine a perceptual visual quality score for the image, and adjusting operation of the decoder based upon the perceptual visual quality score. A decoding system includes a memory to receive wavelet-transformed compressed image data having a predetermined number of subbands, and a decoder to determine a number of wavelet coefficients in a subset of the predetermined number of subbands, measure an amount of blur in the image data based upon the number of wavelet coefficients, and adjust a decoding process based upon the amount of blur.
1. A method of detecting image quality in a wavelet transform-encoded image, comprising:
receiving wavelet transform-encoded image data partitioned into tiles at a decoder, each tile partitioned into a number of subbands;
evaluating, for each tile, a number of wavelets in a subset of the number of subbands, the subset consisting of fewer than the number of subbands;
assigning a measure to each of the subbands evaluated for each tile; and
using the measures to determine a perceptual visual quality score for the wavelet transform-encoded image data, by combining the measure for each of the subbands of the subset of the number of subbands into the perceptual visual quality score, wherein the determined perceptual visual quality score corresponds to an amount of blur;
wherein assigning a measure to each of the subbands evaluated for each tile comprises:
determining, for each pixel in the subband, whether the pixel value is less than a low threshold or higher than a high threshold and incrementing a coefficient counter when the pixel value is less than the low threshold or higher than the high threshold, and
assigning the measure for the subband as a ratio of the coefficient counter value to the number of pixels in the subband.
2. The method of claim 1 , further comprising assigning a weight to each tile, wherein the weight is based upon a relative importance of the tile in the wavelet transform-encoded image data.
3. The method of claim 1 , wherein each tile is partitioned into four subbands.
4. The method of claim 3 , wherein evaluating, for each tile, the number of wavelets in the subset of the number of subbands comprises evaluating the number of wavelets in three of the four subbands.
5. The method of claim 1 , wherein evaluating, for each tile, the number of wavelets in the subset of the number of subbands comprises evaluating the number of wavelets in a luminance component of the wavelet transform-encoded image data.
6. The method of claim 1 , wherein evaluating, for each tile, the number of wavelets in the subset of the number of subbands comprises evaluating the number of wavelets in both a luminance and a chroma component of the wavelet transform-encoded image data.
7. The method of claim 1 , wherein evaluating, for each tile, the number of wavelets in the subset of the number of subbands comprises evaluating the wavelet transform-encoded image data occurring in a resolution layer lower than the final resolution layer used by the decoder to generate the final decoded image.
8. The method of claim 1 , wherein evaluating, for each tile, the number of wavelets in the subset of the number of subbands comprises evaluating the wavelet transform-encoded image data occurring in each of several resolution layers.
9. The method of claim 1 , wherein evaluating, for each tile, the number of wavelets in the subset of the number of subbands is performed only on those tiles that correspond to a foreground region of the wavelet transform-encoded image data.
10. A decoding system, comprising:
a memory to receive wavelet-transformed compressed image data having a predetermined number of subbands; and
a decoder to:
determine a number of wavelet coefficients in a subset of the predetermined number of subbands; and
measure an amount of blur in the wavelet-transformed compressed image data based upon the number of wavelet coefficients; and
wherein measuring an amount of blur in the wavelet-transformed compressed image data based upon the number of wavelet coefficients comprises:
determining, for each pixel in each subband of the subset of the predetermined number of subbands, whether the pixel value is less than a low threshold or higher than a high threshold and incrementing a respective subband coefficient counter when the pixel value is less than the low threshold or higher than the high threshold,
calculating a subband score for each subband of the subset of the predetermined number of subbands, wherein the subband score for each subband is a ratio of the respective subband coefficient counter value to the respective number of pixels in the subband, and
combining the subband scores for each subband of the subset of the predetermined number of subbands into the measured amount of blur.