Noise Removal in Embedded Image With Bit Approximation

Stego-images are often contaminated by interchannel noise or active attack when communicating on the Web. And it is challenging to restore embedded image from corrupted stego-image. This paper studies a k NN-bit approximation algorithm remove noises in image. The proposed distinguishes reliable bits extracted bits, and estimates pixel values keeping unchanged correcting unreliable bits. Specifically, 8th (highest) bit of can be approximated with its nearest neighbor pixels. then, if an locates at any one $5^{th}\sim 7^{th}$ pixel, adjusted two neighbors where in-between these neighbors. Finally, for other each maximum minimum possible pixel. We conduct experiments illustrating efficiency, demonstrate that recover images good visual quality stego-images.