A method to convert .tif files to .csv files for rapidly accessing pixel values.

I am proceesing data from 35,000 tif files, 96 at a time. Reading 96 .tif files into RAM (650 1500 pixels each) is very slow, and quickly bogs down (memory leak?). A simpler way to go seems to be to extract the pixel values individually from the files (using csvread(filename,row,column,range)), which avoids cramming the files into memory. Is there a quick way to convert .tif to .csv files? Alternatively, is there a simple way to acess specific pixels from a tif without reading the file inot memory?

[Image Analyst]

Explain exactly what "650 1500 pixels each" means. Do you mean "650 by 1500"? Or something else. It doesn't make sense as it is.

[Joseph Cheng]

if you know your image size (650x1500?) then why go the CSV route and go RAW binary file? This should save file sizes and space as you're not "wasting" bytes for commas+\n+\r and each pixel value should fit nicely in a byte or two and not spread over multiple ascii characters.

Additional benefits if you are going to do pixel comparisons then instead of CSV files and scanning to the particular row, read in that line and parsing out the whole row, just do some quick math to read pixel values at 650*(row-1)*bytes_per_pix+bytes_per_pix*column.

[per isakson]

• How do you read the tiff-files?
• Do you use the Tiff class?
• How much RAM?

[dpb]

Well, 96 images are almost 90 MB if they're 8-bit or less; otherwise multiply by the number of bytes required/pixel (3 for 24-bit, for example).

Storing a 16-bit ASCII representation of a 24-bit number will expand that to being 8-bytes/pixel to represent all values--hardly a memory tradeoff.

And, what is "extract the pixel values individually from the files.." other than reading into memory? How are you going to do anything with the files other than reading them into memory?

There is the special syntax in imread for TIFF files of

        Parameter name   Value
        --------------   -----
         ....

        'PixelRegion'    {ROWS, COLS}.  imread returns the sub-image
                         specified by the boundaries in ROWS and COLS.  ROWS
                         and COLS must be either two- or three-element
                         vectors.  If two elements are provided, they denote
                         the 1-based indices [START STOP].  If three elements
                         are provided, the indices [START INCREMENT STOP]
                         allow image downsampling.