图像分析
一、本节提供的函数描述，通常应用于图像分析应用。
图像边界和周长函数
边界和周界函数，即IMG_boundary和IMG_perimetcr两个函数。它们通常在结构视觉应用中作为结构算子。
形态学操作函数
IMG_dilate_bin和IMG_erode_bin两个函数是图像学算子，通常用于提高二进制图像膨胀和二进制图像侵蚀算法效果。膨胀和侵蚀在图像处理操作中具有基础的意义，比如开和闭操作都可以从膨胀和侵蚀中建立起来。这些函数在机器视觉和医学成像方面非常有用。
边缘检测函数
在机器视觉系统中通常使用边界检测技术。在许多算法中都存在边界检测技术，最通用的是Sobel边界检测。IMG_sobel子程序提供了一个边界检测算法优化执行的子程序。
图像直方图函数
直方图用来生成图像的柱状图。图像的直方图是一个图像亮度级的统计。例如，对于一个8位像素亮度级别的灰度图像，直方图将包括对应可能的256个像素亮度的256bins。每一个bin包含图像中像素点的个数，尤其是亮度值。
图像二值化函数
在图像和视频处理系统中图像阀值操作的不同形式满足不同的图像处理需求。例如，一种阀值可以用于把灰度图像数据转化为二进制图像数据，以用于二进制形态处理。另一个阀值可以用于剪裁图像数据以便得到渴望的范围。在机器视觉应用中，阀值用于简单的分割。
- IMG_clipping_16s
- IMG_thr_gt2max_8
- IMG_thr_gt2thr_8
- IMG_thr_le2min_8
- IMG_thr_le2thr_8
- IMG_thr_gt2max_16
- IMG_thr_gt2thr_16
- IMG_thr_le2min_16
- IMG_thr_le2thr_16
二、本节提供的函数描述，通常应用于图像滤波与格式转换。图像滤波与格式转换
颜色空间转换函数
IMG_ycbcr422p_rgb565子程序实现图像从YCbCr格式到RGB格式的转化，以实现在MPEG和JPEG解码系统中的视频数据在RGB显示器中播放。
下列子程序开辟一个little endian或big endian格式的YcrYCb彩色图像缓冲区。
These routines take a packed YCrYCb colorbuffer in big-endian or little-endian format and expand the constituent colorelements into separate buffers in little-endian byte ordering.
图像卷积函数
卷积函数用于普通图像3X3、5x5、7x7和11x11的滤波，比如图像平滑和锐化处理方法等。
- IMG_conv_11x11_i16s_c16s
- IMG_conv_11x11_i8_c8s
- IMG_conv_7x7_i16s_c16s
- IMG_conv_7x7_i8_c16s
- IMG_conv_7x7_i8_c8s
- IMG_conv_5x5_i16s_c16s
- IMG_conv_5x5_i8_c16s
- IMG_conv_5x5_i8_c8s
- IMG_conv_3x3_i16s_c16s
- IMG_conv_3x3_i16_c16s
- IMG_conv_3x3_i8_c8s
图像相关性函数
相关性函数用于图像匹配操作。在机器视觉、医学成像和安防方面，图像匹配处理是非常有用的。函数实现对3x3、5x5、11x11像素区域高度优化相关性处理。
- IMG_corr_3x3_i8_c8
- IMG_corr_3x3_i8_c16s
- IMG_corr_3x3_i16s_c16s
- IMG_corr_3x3_i16_c16
- IMG_corr_5x5_i16s_c16s
- IMG_corr_11x11_i8_c16s
- IMG_corr_11x11_i16s_c16s
库中提供了两个相关性函数IMG_corr和IMG_corr_gen。IMG_corr_gen是一个更普通的版本，能够对用户特殊的像素区域大小(在文档约束之内)进行相关性处理。IMG_corr_gen_iq函数处理32bitQ-point数据。
误差扩散函数
二进制值输出误差扩散技术广泛用在印花行业中。最广泛应用的误差扩散算法是Floyd-Steinberg算法。在这个函数中对该算法进行了优化。
中值滤波函数
在图像恢复处理中，中值滤波用于使成像中的脉冲噪音产生的影响降到最低。该处理方法可以保护脉冲影响十分严重的局部区域，在安防、机器视觉和视频压缩系统中得到广泛的应用。IMG_median_3x3函数中值滤波提供了优化实现。
像素扩展函数
IMG_pix_expand子程序用于通过零扩展技术把8位像素点扩展到16位；IMG_pix_sat用于把16位有符号数转换成8位无符号数。通常用于处理其他子程序的输入和输出子程序，例如水平和垂直缩放子程序。
图像压缩与解压
三、本节提供的函数描述，通常应用于标准JPEG、MPEG video和H.26x的压缩/解压
正和逆离散余弦变换函数
IMGLIB provides forwardand inverse DCT (Discrete Cosine Transform) functions. These functions areapplicable for a wide range of compression standards such as JPEGEncode/Decode, MPEG Video Encode/Decode, and H.26x Encode/Decode. Thesecompression standards are used in diverse end-applications:
- JPEG is used in printing, photography, security systems, etc.
- MPEG video standards are used in digital TV, DVD players, set-top boxes, video-on-demand systems, video disc applications, multimedia/streaming media applications, etc.
- H.26x standards are used in video telephony as well as some streaming media applications
Note that the inverse DCT functionperforms an IEEE 1180-1990 compliant inverse DCT, including rounding andsaturation to signed 9-bit quantities. The forward DCT rounds the output valuesfor improved accuracy. These factors can have significant effect on the finalresult in terms of picture quality, and are important to consider whenimplementing DCT-based systems or comparing the performance of differentDCT-based implementations.
高性能运动估计函数
The following functionsare provided to enable high performance motion estimation algorithms that areused in applications such as MPEG Video Encode or H.26x Encode.
Video encoding is useful invideo-on-demand systems, streaming media systems, video telephony, etc. Motionestimation is typically one of the most computationally intensive operations invideo encoding systems; the provided functions are highly optimized tosignificantly improve such systems.
量化函数
Quantization is anintegral step in many image/video compression systems, including those based onwidely used variations of DCT-based compression such as JPEG, MPEG, and H.26x.The function provided below may be used in such systems to perform the requiredquantization step.
小波处理函数
Wavelet processing isused in emerging standards such as JPEG2000 and MPEG-4, where it is typicallyused to provide highly efficient still picture compression. Various proprietaryimage compression systems are also wavelet-based. This release includes theutilities IMG_wave_horz and IMG_wave_vert for computing horizontal and verticalwavelet transforms. Together, they can compute 2-D wavelet transforms for imagedata. The routines are flexible enough, within documented constraints, toaccommodate a wide range of specific wavelets and image dimensions.