java使用mozjpeg高效无损压缩jpeg

这些年来，网站的图片使用数量和大小都在与日俱增，而HTML、JS和CSS文件大小都相对减小了。也就是说在页面加载的过程中，图片占用了大量的网络流量。所以减小图片的大小可以显著优化页面加载速度。

jpeg/jpg 图片在网站使用中比较常见，可以使用java自带的 java.awt.Image 处理，或者用第三方图片jar（如google的开源工具thumbnailator），今天介绍一个jpeg的处理库MozJPEG。

说起MozJPEG，首先了解一下jpeg的发展历史：

（1）libjpeg：开发时间最早，使用最广泛的 JPEG 库。由于 JPEG 标准过于复杂和模糊，并没有其他人去实现，所以这个库是 JPEG 的事实标准；
（2）libjpeg-turbo：一个致力于提升编解码速度的 JPEG 库。它基于 libjpeg 进行了改造，用 SIMD 指令集 (MMX、SSE2、NEON) 重写了部分代码，官网称相对于 libjpeg 有 2 到 4 倍的性能提升；
（3）MozJPEG：一个致力于提升压缩比的 JPEG 库。它是 Mozilla 在 2014 年发布的基于 libjpeg-turbo 进行改造的库，相对于 libjpeg 有 5% ~ 15% 的压缩比提升，但相应的其编码速度也慢了很多。

虽然说他的速度慢了一些，但比guetzli快了很多，个人觉得更适合web的图片处理。

话不多言，先对比一下压缩前后的图片。如下图。

怎么样，很惊讶吧！色彩和亮度跟原图差不多，也不模糊，感觉称得上无损压缩了，上图是使用了80的压缩质量。压缩质量值可以自己控制。当然了，压缩率设置越低的话，图片也会越模糊。

官方的代码中，有java的测试代码，https://github.com/mozilla/mozjpeg/tree/master/java

java调用mozjpeg，主要是用 JNI

static void load() {
    try {
        System.loadLibrary("turbojpeg");
    } catch (UnsatisfiedLinkError e) {
        String os = System.getProperty("os.name").toLowerCase();
        if (os.indexOf("mac") >= 0) {
            try {
                System.load("@CMAKE_INSTALL_FULL_LIBDIR@/libturbojpeg.jnilib");
            } catch (UnsatisfiedLinkError e2) {
                System.load("/usr/lib/libturbojpeg.jnilib");
            }
        } else {
            try {
                System.load("@CMAKE_INSTALL_FULL_LIBDIR@/libturbojpeg.so");
            } catch (UnsatisfiedLinkError e3) {
                String libdir = "@CMAKE_INSTALL_FULL_LIBDIR@";
                if (libdir.equals("@CMAKE_INSTALL_DEFAULT_PREFIX@/lib64")) {
                    System.load("@CMAKE_INSTALL_DEFAULT_PREFIX@/lib32/libturbojpeg.so");
                } else if (libdir.equals("@CMAKE_INSTALL_DEFAULT_PREFIX@/lib32")) {
                    System.load("@CMAKE_INSTALL_DEFAULT_PREFIX@/lib64/libturbojpeg.so");
                } else {
                    throw e3;
                }
            }
        }
    }
}

static void load() {

try {

System.loadLibrary("turbojpeg");

} catch (UnsatisfiedLinkError e) {

String os = System.getProperty("os.name").toLowerCase();

if (os.indexOf("mac") >= 0) {

try {

System.load("@CMAKE_INSTALL_FULL_LIBDIR@/libturbojpeg.jnilib");

} catch (UnsatisfiedLinkError e2) {

System.load("/usr/lib/libturbojpeg.jnilib");

}

} else {

try {

System.load("@CMAKE_INSTALL_FULL_LIBDIR@/libturbojpeg.so");

} catch (UnsatisfiedLinkError e3) {

String libdir = "@CMAKE_INSTALL_FULL_LIBDIR@";

if (libdir.equals("@CMAKE_INSTALL_DEFAULT_PREFIX@/lib64")) {

System.load("@CMAKE_INSTALL_DEFAULT_PREFIX@/lib32/libturbojpeg.so");

} else if (libdir.equals("@CMAKE_INSTALL_DEFAULT_PREFIX@/lib32")) {

System.load("@CMAKE_INSTALL_DEFAULT_PREFIX@/lib64/libturbojpeg.so");

} else {

throw e3;

}

来跑一下官方例子：

1. 安装libjpeg-turbo

我是在 mac 笔记本运行的，先安装libjpeg-turbo，可以使用brew。

brew install jpeg-turbo

1	brew install jpeg-turbo

2. 指定库文件路径

比如我的安装目录在/opt/libjpeg-turbo，运行TJExample.java文件时，加上vm参数，指定库文件路径，如下

-Djava.library.path=/opt/libjpeg-turbo/lib

1	-Djava.library.path=/opt/libjpeg-turbo/lib

也可以在代码上指定库文件路径，不同的操作系统，库文件格式不一样：mac是dylib格式，windows是dll格式，linux是so格式。

System.loadLibrary("/opt/libjpeg-turbo/lib/libturbojpeg.dylib");

1	System.loadLibrary("/opt/libjpeg-turbo/lib/libturbojpeg.dylib");

最后，运行结果如下

USAGE: java [Java options] TJExample <Input image> <Output image> [options]

Input and output images can be in any image format that the Java Image I/O
extensions understand.  If either filename ends in a .jpg extension, then
the TurboJPEG API will be used to compress or decompress the image.

Compression Options (used if the output image is a JPEG image)
--------------------------------------------------------------

-subsamp <444|422|420|gray> = Apply this level of chrominance subsampling when
     compressing the output image.  The default is to use the same level of
     subsampling as in the input image, if the input image is also a JPEG
     image, or to use grayscale if the input image is a grayscale non-JPEG
     image, or to use 4:4:4 subsampling otherwise.

-q <1-100> = Compress the output image with this JPEG quality level
     (default = 95).

Decompression Options (used if the input image is a JPEG image)
---------------------------------------------------------------

-scale M/N = Scale the input image by a factor of M/N when decompressing it.
(M/N = 2/1, 15/8, 7/4, 13/8, 3/2, 11/8, 5/4, 9/8, 1/1, 7/8, 3/4, 5/8, 1/2, 3/8, 1/4, or 1/8)

-hflip, -vflip, -transpose, -transverse, -rot90, -rot180, -rot270 =
     Perform one of these lossless transform operations on the input image
     prior to decompressing it (these options are mutually exclusive.)

-grayscale = Perform lossless grayscale conversion on the input image prior
     to decompressing it (can be combined with the other transform operations
     above.)

-crop WxH+X+Y = Perform lossless cropping on the input image prior to
     decompressing it.  X and Y specify the upper left corner of the cropping
     region, and W and H specify the width and height of the cropping region.
     X and Y must be evenly divible by the MCU block size (8x8 if the input
     image was compressed using no subsampling or grayscale, 16x8 if it was
     compressed using 4:2:2 subsampling, or 16x16 if it was compressed using
     4:2:0 subsampling.)

General Options
---------------

-display = Display output image (Output filename need not be specified in this
     case.)

-fastupsample = Use the fastest chrominance upsampling algorithm available in
     the underlying codec.

-fastdct = Use the fastest DCT/IDCT algorithms available in the underlying
     codec.

-accuratedct = Use the most accurate DCT/IDCT algorithms available in the
     underlying codec.

USAGE: java [Java options] TJExample <Input image> <Output image> [options]

Input and output images can be in any image format that the Java Image I/O

extensions understand. If either filename ends in a .jpg extension, then

the TurboJPEG API will be used to compress or decompress the image.

Compression Options (used if the output image is a JPEG image)

--------------------------------------------------------------

-subsamp <444|422|420|gray> = Apply this level of chrominance subsampling when

compressing the output image. The default is to use the same level of

subsampling as in the input image, if the input image is also a JPEG

image, or to use grayscale if the input image is a grayscale non-JPEG

image, or to use 4:4:4 subsampling otherwise.

-q <1-100> = Compress the output image with this JPEG quality level

(default = 95).

Decompression Options (used if the input image is a JPEG image)

---------------------------------------------------------------

-scale M/N = Scale the input image by a factor of M/N when decompressing it.

(M/N = 2/1, 15/8, 7/4, 13/8, 3/2, 11/8, 5/4, 9/8, 1/1, 7/8, 3/4, 5/8, 1/2, 3/8, 1/4, or 1/8)

-hflip, -vflip, -transpose, -transverse, -rot90, -rot180, -rot270 =

Perform one of these lossless transform operations on the input image

prior to decompressing it (these options are mutually exclusive.)

-grayscale = Perform lossless grayscale conversion on the input image prior

to decompressing it (can be combined with the other transform operations

above.)

-crop WxH+X+Y = Perform lossless cropping on the input image prior to

decompressing it. X and Y specify the upper left corner of the cropping

region, and W and H specify the width and height of the cropping region.

X and Y must be evenly divible by the MCU block size (8x8 if the input

image was compressed using no subsampling or grayscale, 16x8 if it was

compressed using 4:2:2 subsampling, or 16x16 if it was compressed using

4:2:0 subsampling.)

General Options

---------------

-display = Display output image (Output filename need not be specified in this

case.)

-fastupsample = Use the fastest chrominance upsampling algorithm available in

the underlying codec.

-fastdct = Use the fastest DCT/IDCT algorithms available in the underlying

codec.

-accuratedct = Use the most accurate DCT/IDCT algorithms available in the

underlying codec.

以下网站的jpeg图片，就是用了mozjpeg压缩的，可以前往测试效果

demo网站

发表评论 取消回复

发表评论取消回复