A new illumination estimation method based on local gradient for retinex

In this paper, we present the new illumination estimation method for retinex algorithm, which is the color constancy algorithm under non-uniform illumination condition. In this algorithm, the most important factor is illumination estimation. Because properly estimated illumination guarantees good enhanced image. There are many algorithms for illumination estimation, LPF, MSR, bilateral filter, and so on. They have merits for illumination estimation, but they have demerits too. Proposed algorithm has edge preserving property that LPF, MSR haven´t. Proposed algorithm also has fast processing time using the fast algorithm. In this paper, we compare proposed algorithm with previous illumination estimation algorithms. And we show proposed algorithm superior to previous algorithms through various experiments.