Abstract: Because human eyes may become less sensitive to dark areas around very bright areas in a video (known as glare masking), we may use coarser quantization in such dark areas. Considering the extra distortion that human eyes can tolerate in a block with glare masking, a quantization ratio for the block can be calculated. The quantization parameter for the block can then be scaled up using the quantization ratio to form an adjusted quantization parameter. In one embodiment, the adjusted quantization parameter can be derived at the decoder side, and thus, transmission of quantization ratio information is not needed. In particular, we can estimate the luminance of a current block based on a predicted block and de-quantized DC coefficient, and use the luminance of causal neighboring blocks and the estimated luminance of the current block to estimate the adjusted quantization parameter.

Abstract: Because human eyes may become less sensitive to dark areas around very bright areas in a video (known as glare masking), we may use coarser quantization in such dark areas. Considering the extra distortion that human eyes can tolerate in a block with glare masking, a quantization ratio for the block can be calculated. The quantization parameter for the block can then be scaled up using the quantization ratio to form an adjusted quantization parameter. In one embodiment, the adjusted quantization parameter can be derived at the decoder side, and thus, transmission of quantization ratio information is not needed. In particular, we can estimate the luminance of a current block based on a predicted block and de-quantized DC coefficient, and use the luminance of causal neighboring blocks and the estimated luminance of the current block to estimate the adjusted quantization parameter.