| 摘要: |
| 目的 针对当前较多图像修复算法难以根据不同纹理结构来自适应调整修复块的尺寸，导致修复结果中存在不连续效应和模糊效应等不足，提出一种结合匹配调节法则和梯度约束模型的图像修复算法。方法 首先，利用平滑因子对置信度项进行约束，构造优先级判定模型，对待修复块的优先级进行度量，确定优先修复块。随后，通过SSD模型度量样本块之间的匹配结果，并根据匹配结果制定匹配调节法则，使得样本块能根据匹配度自适应调节其大小，以提高修复质量。最后，将梯度模值块中像素点的均方距离度量结果与样本块中像素点相结合，构造梯度约束模型，用以获取最佳匹配块对待修复块进行填充修复。结果 实验结果表明，与当前图像修复算法相比，该算法修复的图像具有更好的修复质量，在像素丢失率较高的情况下，仍然具有较高的相似度值。结论 所提算法具备较好的修复视觉质量，可用于被大面积损坏图像的修复。 |
| 关键词: 图像修复 优先级判定模型 匹配调节法则 SSD 梯度约束模型 平滑因子 |
| DOI：10.19554/j.cnki.1001-3563.2018.13.038 |
| 分类号:TP391 |
| 基金项目:江苏省自然科学基金（K2014401） |
|
| Image Inpainting Algorithm Based on Matching Adjustment Rule and Gradient Constraint Model |
|
WU Yin-fang1, ZHU Sen-cheng2
|
|
1.Jiangsu College of Nursing, Huai'an 223300, China;2.Nanjing Medical University, Nanjing 210029, China
|
| Abstract: |
| The work aims to propose an image inpainting algorithm based on matching adjustment rule and gradient constraint model, regarding such deficiencies as discontinuity effect and fuzzy effect in the restoration results caused by the situation that currently it is hard for major image inpainting algorithms to make self-adaptive adjustments of the restored block size according to different texture structures. Firstly, the confidence item was restrained by smoothing factor. The priority decision model was constructed to measure the priority of the block to be restored and determine the preferably restored block. Then, the SSD model was used to measure the matching results between sample blocks, and the matching adjustment rule was set out according to the matching results, so that the adaptive adjustment of sample block size could be made according to the matching degree, in order to improve the restoration quality. Finally, the gradient constraint model was constructed by combining the measurements of mean square distance of the pixels in the gradient modulus block with the pixels in the sample blocks, so as to obtain the best matching blocks for the filling and restoration of the blocks to be restored. The experimental results showed that, compared with the current image inpainting algorithm, the image restored by the proposed algorithm had better restoration quality and still had higher similarity value when the pixel loss rate was higher. The proposed algorithm has better visual restoration quality and can be used for the restoration of extensively damaged image. |
| Key words: image inpainting priority decision model matching adjustment rule SSD gradient constraint model smoothing factor |
渝公网安备 50010702501716号