This article presents a fast approximation method for eficient rendering of multi-fragment efiects via bucket sort on GPU. In particular, a bucket array of size K is allocated per pixel location using MRT as storage, and the depth range of each pixel is consequently divided into K subintervals. During rasterization, fragments within the kth (k = 0, 1, …, K - 1) subinterval will be routed to the kth bucket by a bucket sort. Collisions will happen when multiple fragments are routed to the same bucket, which can be alleviated by multi-pass approach or an adaptive scheme. Our algorithm shows great speedup to the classical depth peeling with visually faithful results, especially for large scenes with high complexity.

;

This article presents a fast approximation method for eficient rendering of multi-fragment efiects via bucket sort on GPU. In particular, a bucket array of size K is allocated per pixel location using MRT as storage, and the depth range of each pixel is consequently divided into K subintervals. During rasterization, fragments within the kth (k = 0, 1, …, K - 1) subinterval will be routed to the kth bucket by a bucket sort. Collisions will happen when multiple fragments are routed to the same bucket, which can be alleviated by multi-pass approach or an adaptive scheme. Our algorithm shows great speedup to the classical depth peeling with visually faithful results, especially for large scenes with high complexity.