Learn how easy it is to sync an existing GitHub or Google Code repo to a SourceForge project!

## Re: [Vxl-users] vil fft question

 Re: [Vxl-users] vil fft question From: Ian Scott - 2005-12-02 10:35:20 ```Timothy Chan wrote: > Ian, > > Thanks for all your help and putting up with my stupid questions. I have > successfully used both fft fwd and bwd now on a number of images so that has > been a great help. > > There is one file in particular that I cannot seem to run the fft on, and > when I do I get the message "assertion failure '!'you probably gave a signal > size not of the form 2^p 3^q 5^r" Can you explain to me what this means and > how I can get around it? If there was somebody here that I could ask I > certain would, but unfortunately there isn't. We don't mind questions about VXL - its just non-VXL programming and tool-chain management questions we would rather avoid. However, your questions would be better sent to vxl-users. I'm not an expert on the FFT code, but at a guess the size of your image needs to be (2^pi*3^qi*5^ri,2^pj*3^qj*5^rj), where all the variables are non-negative integers. Most FFT algorithms require your image to be sized as a power of 2, in order to have a Fast Fourier Transform rather than just a Fourier Transform. VXL's FFT appears to be rather generous in allowing factors of 3 and 5 as well. I suspect the solution is to pad your image with some values so that it is the next appropriate size. Some of the links at http://www.google.com/search?q=padding+data+for+FFT appear to suggest that padding with 0 is generally acceptable. Ian. If your image is > > -Tim Chan > ```

 Re: [Vxl-users] vil fft question From: Ian Scott - 2005-12-02 10:35:20 ```Timothy Chan wrote: > Ian, > > Thanks for all your help and putting up with my stupid questions. I have > successfully used both fft fwd and bwd now on a number of images so that has > been a great help. > > There is one file in particular that I cannot seem to run the fft on, and > when I do I get the message "assertion failure '!'you probably gave a signal > size not of the form 2^p 3^q 5^r" Can you explain to me what this means and > how I can get around it? If there was somebody here that I could ask I > certain would, but unfortunately there isn't. We don't mind questions about VXL - its just non-VXL programming and tool-chain management questions we would rather avoid. However, your questions would be better sent to vxl-users. I'm not an expert on the FFT code, but at a guess the size of your image needs to be (2^pi*3^qi*5^ri,2^pj*3^qj*5^rj), where all the variables are non-negative integers. Most FFT algorithms require your image to be sized as a power of 2, in order to have a Fast Fourier Transform rather than just a Fourier Transform. VXL's FFT appears to be rather generous in allowing factors of 3 and 5 as well. I suspect the solution is to pad your image with some values so that it is the next appropriate size. Some of the links at http://www.google.com/search?q=padding+data+for+FFT appear to suggest that padding with 0 is generally acceptable. Ian. If your image is > > -Tim Chan > ```