Method of applying edge offset

摘要

Provided is a method generates an edge index of a current sample, and applies an edge offset corresponding to the edge index to the current sample. The edge index is generated using the differences between a current sample and two neighboring samples determined by an edge offset type. Accordingly, the difference between original samples and reconstructed samples are effectively reduced by generating the optimum edge index. Also, the quantity of bits required for reducing the differences are reduced by fixing the sign of offset to positive or negative.

主权项

1. A method of post-processing a reconstructed image, the method comprising:
determining a boundary strength for each 4-sample edge which is a prediction edge or a transform edge and lies on an 8×8 sample grid; determining whether deblocking filtering is applied on the 4-sample edge or not using the boundary strength and a boundary quantization parameter; filtering the 4-sample edge if the deblocking filtering is applied on the 4-sample edge; and applying an edge offset if a sample adaptive offset (SAO) type indicates an edge offset, wherein applying the edge offset comprises:
generating an edge index of a current sample; andapplying an edge offset corresponding to the edge index to the current sample,wherein the edge index is generated using the following equation,
edgeIdx=2+sign3(recPicture(x)−recPicture(x−1))+sign3(recPicture(x)−recPicture(x+1))wherein the function sign3(y) is equal to 1 if the y is larger than 0, the function sign3(y) is equal to −1 if the y is smaller than 0, and the function sign3(y) is equal to 0 if the y is equal to 0, andwherein variable recPicture(x) represents the current sample value, the variables recPicture(x−1) and recPicture(x+1) represent the two neighboring sample values, and wherein the boundary quantization parameter is an average of a quantization parameter of a P block containing sample p0 and a quantization parameter of a Q block containing sample q0,
wherein the quantization parameter of the P block is generated by using a quantization parameter predictor and a differential quantization parameter of the P block, and, if two or more quantization parameters are available among a left quantization parameter, an above quantization parameter and a previous quantization parameter of the P block, the quantization parameter predictor of the P block is generated using two available quantization parameters which are determined according to a predetermined order.