最小二乘估计耦合小波相似性的边缘检测算法

为解决当前图像边缘提取时存在边缘不连续、虚假边缘及边缘定位不准确等不足,设计最小二乘估计耦合小波相似性的不均匀光照边缘检测算法。通过将图像分割成一系列的子块,计算其相邻像素的相似映射,平均不均匀光照强度;为获得边缘信息,计算相似映射下图像邻域像素和方向小波滤波器之间的相似度;将加权最小二乘估计正则化结合小波相似度来处理光照差异,提供亮度不变的相似性和精确的边缘定位与提取,获取光照校正后的水平、垂直和对角子带;通过合成所有子带与相似映射获得边缘。实验结果表明,与当前边缘检测方案相比,在光照不均匀条件下,所提方法具有更高的边缘提取质量。

Edge detection algorithm based on least squares estimation and wavelet similarity

To solve the problems as edge discontinuity,false edge and inaccurate edge location in current image edge extraction methods,an uneven illumination edge detection scheme based on least squares estimation and wavelet similarity was designed.The image was divided into a series of sub-blocks to calculate the similarity mapping of its adjacent pixels,and the uneven illumination intensity was better averaged through similarity map.To obtain edge information,the similarity between neighborhood pixels and directional wavelet filters was calculated.The weighted least squares estimation regularization was combined with wavelet similarity to deal with the illumination difference for providing brightness-invariant similarity and accurate edge location and extraction to obtain horizontal,vertical and diagonal subbands after illumination correction.The edges were obtained by synthesizing all subbands and similar maps.Experimental data show that under the condition of non-uniform illumination,the proposed method has higher quality of edge extraction than the current advanced edge extraction methods.