基于背景先验点的Photoshop交互式图像自动降噪处理研究

Research on interactive Photoshop images automatic denoising based on background prior points

为了降低Photoshop图像交互过程中产生的噪点,改善图像清晰度,提出一种基于背景先验点的Photoshop交互式图像自动降噪处理方法。计算输入图像的每个超像素显著值,分割非重叠小区域,保证每个区域都具有显著值;通过区域显著值确定Photoshop交互式图像背景先验点,为图像自动降噪提供位置。计算Photoshop交互式图像的分布方差,考虑噪声模型加1修正,获取修正后的图像噪声模型;再寻找图像相似块,计算任意两个小片a、b之间的欧氏距离,将噪声图像中的噪声点宽度实际极值对应的位置设定为背景先验点,避免陷入局部极值点。最后,对像素矩阵进行3D变换处理,获取降噪3D块像素矩阵,按照降噪图像匹配相似块的方法匹配含噪图像的相似块,经过二次降噪后得到最终降噪图像。实验结果表明,所提方法最小峰值信噪比为42 dB,获取的降噪图像无噪点且整体清晰,视觉效果有明显提高。

In order to reduce the noise generated during Photoshop image interaction and improve image clarity,a method of photoshop interactive image automatic denoising based on background prior points is proposed.The saliency value of each superpixel in the input image is calculated,and non overlapping small regions are segmented,so as to ensure that each region has saliency values.The background prior points of Photoshop interactive images are determined by means of regional saliency values to provide a location for automatic image denoising.The distribution variance of the interactive Photoshop image is calculated,and the modified image noise model is obtained by considering noise model with 1 correction.The similar blocks in the image is found,the Euclidean distance between any two small pieces a and b,and the position corresponding to the actual set value of the noise point width in the noisy image is set as the background prior point to avoid falling into local extreme points.The 3D transformation is performed on the pixel matrix to obtain a denoised 3D block pixel matrix,similar blocks of noisy images is matched according to the method of matching similar blocks in denoised images,and obtain the final denoised image after secondary denoising.The experimental results show that the minimum peak signal-to-noise ratio of the proposed method is 42 dB,and the obtained denoised image is noise free and overall clear,with a significant improvement in visual effect.