/usr/share/doc/beignet-dev/howto/video-motion-estimation-howto.html is in beignet-dev 1.3.2-2.
This file is owned by root:root, with mode 0o644.
The actual contents of the file can be viewed below.
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 | <!DOCTYPE html>
<html xmlns="http://www.w3.org/1999/xhtml">
<head>
<meta http-equiv="Content-Type" content="text/html; charset=utf-8" />
<title>video-motion-estimation-howto</title>
<meta name="viewport" content="width=device-width, initial-scale=1" />
<link rel="stylesheet" href="../style.css" type="text/css" />
<link rel="stylesheet" href="../local.css" type="text/css" />
</head>
<body>
<div class="page">
<div class="pageheader">
<div class="header">
<span>
<span class="parentlinks">
<a href="../index.html">wiki</a>/
<a href="../index.html">howto</a>/
</span>
<span class="title">
video-motion-estimation-howto
</span>
</span>
</div>
</div>
<div id="pagebody">
<div id="content" role="main">
<h1>Video Motion Vector HowTo</h1>
<p>Beignet now supports cl_intel_accelerator and cl_intel_motion_estimation, which are
Khronos official extensions. It provides a hardware acceleration of video motion
vector to users.</p>
<h2>Supported hardware platform</h2>
<p>Only 3rd Generation Intel Core Processors is supported for vme now. We will consider
to support more platforms if necessary.</p>
<h2>Steps</h2>
<p>In order to use video motion estimation provided by Beignet in your program, please follow
the steps as below:</p>
<ul>
<li><p>Create a cl_accelerator_intel object using extension API clCreateAcceleratorINTEL, like
this:
_accelerator_type_intel accelerator_type = CL_ACCELERATOR_TYPE_MOTION_ESTIMATION_INTEL;
cl_motion_estimation_desc_intel vmedesc = {CL_ME_MB_TYPE_16x16_INTEL,
CL_ME_SUBPIXEL_MODE_INTEGER_INTEL,
CL_ME_SAD_ADJUST_MODE_NONE_INTEL,
CL_ME_SEARCH_PATH_RADIUS_16_12_INTEL
};</p></li>
<li><p>Invoke clCreateProgramWithBuiltInKernels to create a program object with built-in kernels
information, and invoke clCreateKernel to create a kernel object whose kernel name is
block_motion_estimate_intel.</p></li>
<li><p>The prototype of built-in kernel block_motion_estimate_intel is as following:
_kernel void
block_motion_estimate_intel
(
accelerator_intel_t accelerator,
<strong>read_only image2d_t src_image,
</strong>read_only image2d_t ref_image,
<strong>global short2 * prediction_motion_vector_buffer,
</strong>global short2 * motion_vector_buffer,
__global ushort * residuals
);
So you should create related objects and setup these kernel arguments by clSetKernelArg.
Create source and reference image object, on which you want to do video motion estimation.
The image_channel_order should be CL_R and image_channel_data_type should be CL_UNORM_INT8.
Create a buffer object to get the motion vector result. This motion vector buffer representing
a vector field of pixel block motion vectors, stored linearly in row-major order. The elements
(pixels) of this image contain a motion vector for the corresponding pixel block, with its x/y
components packed as two 16-bit integer values. Each component is encoded as a S13.2 fixed
point value(two's complement).</p></li>
<li><p>Use clEnqueueNDRangeKernel to enqueue this kernel. The only thing you need to setup is global_work_size:
global_work_size[0] equal to width of source image, global_work_size[1] equal to height of source
image.</p></li>
<li><p>Use clEnqueueReadBuffer or clEnqueueMapBuffer to get motion vector result.</p></li>
</ul>
<h2>Sample code</h2>
<p>We have developed an utest case of using video motion vector in utests/builtin_kernel_block_motion_estimate_intel.cpp.
Please go through it for details.</p>
<h2>More references</h2>
<p><a href="https://www.khronos.org/registry/cl/extensions/intel/cl_intel_accelerator.txt">https://www.khronos.org/registry/cl/extensions/intel/cl_intel_accelerator.txt</a>
<a href="https://www.khronos.org/registry/cl/extensions/intel/cl_intel_motion_estimation.txt">https://www.khronos.org/registry/cl/extensions/intel/cl_intel_motion_estimation.txt</a>
<a href="https://software.intel.com/en-us/articles/intro-to-motion-estimation-extension-for-opencl">https://software.intel.com/en-us/articles/intro-to-motion-estimation-extension-for-opencl</a></p>
</div>
</div>
<div id="footer" class="pagefooter" role="contentinfo">
<div id="pageinfo">
<div id="backlinks">
Links:
<a href="../Beignet.html">Beignet</a>
</div>
<div class="pagedate">
</div>
</div>
<!-- from wiki -->
</div>
</div>
</body>
</html>
|