使用opencv对图像进行编码,一方面是图像二进制传输的需要,另一方面对图像压缩。以jpeg压缩为例:
1、转为二进制编码
img = cv2.imread(img_path) # 取值范围:0~100,数值越小,压缩比越高,图片质量损失越严重 params = [cv2.IMWRITE_JPEG_QUALITY, ratio] # ratio:0~100 msg = cv2.imencode(".jpg", img, params)[1] msg = (np.array(msg)).tobytes() print("msg:", len(msg))
对于png压缩,改为:
# 取值范围:0~9,数值越小,压缩比越低,图片质量越高 params = [cv2.IMWRITE_PNG_COMPRESSION, ratio] # ratio: 0~9 msg = cv2.imencode(".png", img, params)[1] msg = (np.array(msg)).tobytes()
对于图像解码,使用imdecode即可解为numpy类型图像:
img = cv2.imdecode(np.frombuffer(msg, np.uint8), cv2.IMREAD_COLOR) print(img.shape, type(img))
2、图像质量压缩
原图(48k):
jpg压缩:
img_path = r"E:\img.jpg" img = cv2.imread(img_path) cv2.imwrite(r"E:\ret_80.jpg", img, [cv2.IMWRITE_JPEG_QUALITY, 80]) cv2.imwrite(r"E:\ret_40.jpg", img, [cv2.IMWRITE_JPEG_QUALITY, 40]) cv2.imwrite(r"E:\ret_10.jpg", img, [cv2.IMWRITE_JPEG_QUALITY, 10]) cv2.imwrite(r"E:\ret_0.jpg", img, [cv2.IMWRITE_JPEG_QUALITY, 0])
结果:
压缩后图像大小依次为:49.6K、25.6K、11K、5.02K。jpg压缩明显,压缩到极致时颜色信息损失严重。
png压缩:
img_path = r"E:\img.jpg" img = cv2.imread(img_path) cv2.imwrite(r"E:\ret_0.png", img, [cv2.IMWRITE_PNG_COMPRESSION, 0]) cv2.imwrite(r"E:\ret_3.png", img, [cv2.IMWRITE_PNG_COMPRESSION, 3]) cv2.imwrite(r"E:\ret_6.png", img, [cv2.IMWRITE_PNG_COMPRESSION, 6]) cv2.imwrite(r"E:\ret_9.png", img, [cv2.IMWRITE_PNG_COMPRESSION, 9])
结果:
压缩后图像大小依次为:675K、364K、364K、360K。png格式偏大,压缩率调到最高也还有360K,且成像上无明显变化。
PS:也可以对图像压缩后保存,如:
img_path = r"E:\img.jpg" img = cv2.imread(img_path) params = [cv2.IMWRITE_PNG_COMPRESSION, 0] msg = cv2.imencode(".png", img, params)[1] msg = (np.array(msg)).tobytes() print("msg:", len(msg)) img = cv2.imdecode(np.frombuffer(msg, np.uint8), cv2.IMREAD_COLOR) cv2.imwrite(rr"E:\ret.jpg", img)
bug处理:
早期版本这样写:
msg = (np.array(msg)).tostring() 改为: msg = (np.array(msg)).tobytes() img = cv2.imdecode(np.fromstring(msg, np.uint8), cv2.IMREAD_COLOR) 改为: img = cv2.imdecode(np.frombuffer(msg, np.uint8), cv2.IMREAD_COLOR)
到此这篇关于python cv2图像质量压缩的算法示例的文章就介绍到这了,更多相关python cv2图像质量压缩 内容请搜索服务器之家以前的文章或继续浏览下面的相关文章希望大家以后多多支持服务器之家!
原文链接:https://blog.csdn.net/weixin_34910922/article/details/117537384