论文部分内容阅读
针对图像在传输过程中易引入噪声、色彩质量下降、中值滤波导致图像细节丢失和均值滤波出现模糊等问题,提出了一种可以应用于CMOS图像传感器的图像画质增强和滤波算法。该算法对插值后的Bayer图像数据进行一维空间的增强和降噪处理,首先将图像从RGB空间转换到YUV空间,在Y通道上用改进的直方图均衡化方法实现图像明暗程度的对比度增强调节,对U、V通道采用分段式线性调节方法实现饱和度调节;然后对Y通道进行自适应降噪,对U、V通道进行加权中值滤波降噪,以满足后续处理对图像质量的要求;最后在Y通道上,采用基于Laplace算子的锐化掩模进行锐化处理,保证图像的细节清晰可见。实验结果表明:从图像视觉效果来看,相比单独使用中值和均值滤波,所提出的自适应滤波得到的效果更好,图像细节保存较好、模糊程度低、图像更为清晰,且色彩质量更高。通过对比峰值信噪比(PSNR),对混合噪声进行处理时,该滤波算法的PSNR优于中值和均值滤波,有效地抑制了噪声。整个算法在一维邻域空间进行,更容易在有限的硬件上实现较好的图像处理结果,满足小面积低功耗的要求。
Aiming at the problems such as the image easy to introduce noise, the decline of color quality, the loss of image detail caused by median filtering and the blur of mean filter, this paper proposes an image quality enhancement and filtering algorithm which can be applied to CMOS image sensor. The proposed algorithm performs one-dimensional space enhancement and noise reduction on the interpolated Bayer image data. Firstly, the image is converted from RGB space to YUV space and the contrast enhancement of the brightness of the image is enhanced by the improved histogram equalization method on the Y channel Adjust the saturation adjustment of the U and V channels by using a piecewise linear adjustment method; then, adaptively reduce noise on the Y channel and perform weighted median filtering on the U and V channels to meet the requirements of subsequent processing on the image quality Finally, on the Y-channel, a sharpening mask based on the Laplace operator is used to sharpen the image to ensure that the details of the image are clearly visible. The experimental results show that compared with using median and mean filter separately, the proposed adaptive filtering has better effect, better preservation of image detail, less blur, more clear image, Higher quality. By comparing the peak signal to noise ratio (PSNR), when the mixed noise is processed, the PSNR of the filtering algorithm is better than the median and mean filtering, effectively suppressing the noise. The whole algorithm is carried out in one-dimensional neighborhood space, and it is easier to achieve better image processing results on limited hardware to meet the requirements of small area and low power consumption.