https://community.intel.com/t5/Intel-Integrated-Performance/how-to-adjust-the-alpha-value/m-p/955702#M19053 <DIV>My program need to adjust the alpha value of a image by increasing or decreasing some percent,such as increase 15% alpha,the new value is 1.15*alpha.But which function in IPP can implement it?</DIV> Wed, 02 Jun 2004 01:56:10 GMT pango 2004-06-02T01:56:10Z https://community.intel.com/t5/Intel-Integrated-Performance/how-to-adjust-the-alpha-value/m-p/955702#M19053 <DIV>My program need to adjust the alpha value of a image by increasing or decreasing some percent,such as increase 15% alpha,the new value is 1.15*alpha.But which function in IPP can implement it?</DIV> Wed, 02 Jun 2004 01:56:10 GMT https://community.intel.com/t5/Intel-Integrated-Performance/how-to-adjust-the-alpha-value/m-p/955702#M19053 pango 2004-06-02T01:56:10Z https://community.intel.com/t5/Intel-Integrated-Performance/how-to-adjust-the-alpha-value/m-p/955703#M19054 <P>pango,</P> <P>I'm assuming RGBA and Ipp8u data.</P> <P>You might tryippiMulC_8u_C4RSfs.It will only work if you can express the factor as a fraction with a 2^n denominator. Use {1.0, 1.0, 1.0, factor*scale} as the value.</P> <P>For example, for MulC, you can use {1.0, 1.0, 1.0, 147} with a scaleFactor of 7 (so the scale is 2^7=128). This should multiply the alpha channel by 1.148.</P> <P>You'll have to determine whether the extra 3 multiplications by 1.0 still result in a performance increase. It's not easy to multiply every fourth value efficientlyusing SIMD.</P> <P>I came up with a bunch oftricks like this in the IPP book, although I did find that frequently it took some work and tweaking to get a performance gain.</P> <P>Stewart</P> Fri, 04 Jun 2004 05:30:00 GMT https://community.intel.com/t5/Intel-Integrated-Performance/how-to-adjust-the-alpha-value/m-p/955703#M19054 Stewart_T_Intel 2004-06-04T05:30:00Z