The number of total felling recent questions is a bit too much... This is video steganographic pixel information loss problem, I will be a binary image is embedded video file, method is to read the video frames with XVID codec, carried out in accordance with the 8 by 8 block DCT transformation, there is an embedding strength p, when the embedded pixel 0 the number is negative, otherwise is positive, in order to set the embedding strength p retains as 20, then DCT transform matrix of [2] [2] is set to the position of p, then the inverse transformation to encrypt image, when extracting hidden information will DCT matrix p [2] [2] extracted, if the negative to the pixel is 0, or 1,

But after compression of the reason? Extract the hidden image slightly serious losses, almost can only see about, is my choice of decoder compression too serious? Even if has the embedding strength is set to 20 so high, the output of video can see come out with a grid, video image loss serious, extracted image quality is still not very well,

Should be in a decoder? I should not be embedded by frame or say? If could be embedded into the key frames, and come down to complexity and to improve security, but the opencv did not leave judgment video key frames of interface, it is said that ffmpeg handling of the video is better, but ffmpeg and don't like opencv support so many image processing algorithm, and also difficult to use... Bosses have any better solution?

Hidden information embedded comparison before and after the following