convolution.cpp

2d convolution on image

    // find center position of kernel (half of kernel size)
    kCenter = kSizeY >> 1;                          // center index of vertical kernel
    endIndex = dataSizeY - kCenter;                 // index where full kernel convolution should stop

    // set working pointers
    tmpPtr = tmpPtr2 = tmp;
    outPtr = out;

    // clear out array before accumulation
    for(i = 0; i < dataSizeX; ++i)
        sum[i] = 0;

    // start to convolve vertical direction (y-direction)

    // ROW FROM index=0 TO index=(kCenter-1)
    kOffset = 0;                                    // starting index of partial kernel varies for each sample
    for(i=0; i < kCenter; ++i)
    {
        for(k = kCenter + kOffset; k >= 0; --k)     // convolve with partial kernel
        {
            for(j=0; j < dataSizeX; ++j)
            {
                sum[j] += *tmpPtr * kernelY[k];
                ++tmpPtr;
            }
        }

        for(n = 0; n < dataSizeX; ++n)              // convert and copy from sum to out
        {
            *outPtr = sum[n];                       // store final result to output array
            sum[n] = 0;                             // reset to zero for next summing
            ++outPtr;                               // next element of output
        }

        tmpPtr = tmpPtr2;                           // reset input pointer
        ++kOffset;                                  // increase starting index of kernel
    }

    // ROW FROM index=kCenter TO index=(dataSizeY-kCenter-1)
    for(i = kCenter; i < endIndex; ++i)
    {
        for(k = kSizeY -1; k >= 0; --k)             // convolve with full kernel
        {
            for(j = 0; j < dataSizeX; ++j)
            {
                sum[j] += *tmpPtr * kernelY[k];
                ++tmpPtr;
            }
        }

        for(n = 0; n < dataSizeX; ++n)              // convert and copy from sum to out
        {
            *outPtr = sum[n];                       // store final result to output array
            sum[n] = 0;                             // reset to zero for next summing
            ++outPtr;                               // next output
        }

        // move to next row
        tmpPtr2 += dataSizeX;
        tmpPtr = tmpPtr2;
    }

    // ROW FROM index=(dataSizeY-kCenter) TO index=(dataSizeY-1)
    kOffset = 1;                                    // ending index of partial kernel varies for each sample
    for(i=endIndex; i < dataSizeY; ++i)
    {
        for(k = kSizeY-1; k >= kOffset; --k)        // convolve with partial kernel
        {
            for(j=0; j < dataSizeX; ++j)
            {
                sum[j] += *tmpPtr * kernelY[k];
                ++tmpPtr;
            }
        }

        for(n = 0; n < dataSizeX; ++n)              // convert and copy from sum to out
        {
            *outPtr = sum[n];                       // store final result to output array
            sum[n] = 0;                             // reset to zero for next summing
            ++outPtr;                               // increase ending index of partial kernel
        }

        // move to next row
        tmpPtr2 += dataSizeX;
        tmpPtr = tmpPtr2;                           // next input
        ++kOffset;                                  // increase ending index of kernel
    }
    // END OF VERTICAL CONVOLUTION ////////////////////////

    // deallocate temp buffers
    delete [] tmp;
    delete [] sum;
    return true;
}



///////////////////////////////////////////////////////////////////////////////
// 2D Convolution Fast
// In order to improve the performance, this function uses multple cursors of
// input signal. It avoids indexing input array during convolution. And, the
// input signal is partitioned to 9 different sections, so we don't need to
// check the boundary for every samples.
///////////////////////////////////////////////////////////////////////////////

///////////////////////////////////////////////////////////////////////////////
// unsigned char (8-bit) version
///////////////////////////////////////////////////////////////////////////////
bool convolve2DFast(unsigned char* in, unsigned char* out, int dataSizeX, int dataSizeY,
                    float* kernel, int kernelSizeX, int kernelSizeY)
{
    int i, j, m, n, x, y, t;
    unsigned char **inPtr, *outPtr, *ptr;
    int kCenterX, kCenterY;
    int rowEnd, colEnd;                             // ending indice for section divider
    float sum;                                      // temp accumulation buffer
    int k, kSize;

    // check validity of params
    if(!in || !out || !kernel) return false;
    if(dataSizeX <= 0 || kernelSizeX <= 0) return false;

    // find center position of kernel (half of kernel size)
    kCenterX = kernelSizeX >> 1;
    kCenterY = kernelSizeY >> 1;
    kSize = kernelSizeX * kernelSizeY;              // total kernel size

    // allocate memeory for multi-cursor
    inPtr = new unsigned char*[kSize];
    if(!inPtr) return false;                        // allocation error

    // set initial position of multi-cursor, NOTE: it is swapped instead of kernel
    ptr = in + (dataSizeX * kCenterY + kCenterX); // the first cursor is shifted (kCenterX, kCenterY)
    for(m=0, t=0; m < kernelSizeY; ++m)
    {
        for(n=0; n < kernelSizeX; ++n, ++t)
        {
            inPtr[t] = ptr - n;
        }
        ptr -= dataSizeX;
    }

    // init working  pointers
    outPtr = out;

    rowEnd = dataSizeY - kCenterY;                  // bottom row partition divider
    colEnd = dataSizeX - kCenterX;                  // right column partition divider

    // convolve rows from index=0 to index=kCenterY-1
    y = kCenterY;
    for(i=0; i < kCenterY; ++i)
    {
        // partition #1 ***********************************
        x = kCenterX;
        for(j=0; j < kCenterX; ++j)                 // column from index=0 to index=kCenterX-1
        {
            sum = 0;
            t = 0;
            for(m=0; m <= y; ++m)
            {
                for(n=0; n <= x; ++n)
                {
                    sum += *inPtr[t] * kernel[t];
                    ++t;
                }
                t += (kernelSizeX - x - 1);         // jump to next row
            }

            // store output
            *outPtr = (unsigned char)((float)fabs(sum) + 0.5f);
            ++outPtr;
            ++x;
            for(k=0; k < kSize; ++k) ++inPtr[k];    // move all cursors to next
        }

        // partition #2 ***********************************
        for(j=kCenterX; j < colEnd; ++j)            // column from index=kCenterX to index=(dataSizeX-kCenterX-1)
        {
            sum = 0;
            t = 0;
            for(m=0; m <= y; ++m)
            {
                for(n=0; n < kernelSizeX; ++n)
                {
                    sum += *inPtr[t] * kernel[t];
                    ++t;
                }
            }

            // store output
            *outPtr = (unsigned char)((float)fabs(sum) + 0.5f);
            ++outPtr;
            ++x;
            for(k=0; k < kSize; ++k) ++inPtr[k];    // move all cursors to next
        }

        // partition #3 ***********************************
        x = 1;
        for(j=colEnd; j < dataSizeX; ++j)           // column from index=(dataSizeX-kCenter) to index=(dataSizeX-1)
        {
            sum = 0;
            t = x;
            for(m=0; m <= y; ++m)
            {
                for(n=x; n < kernelSizeX; ++n)
                {
                    sum += *inPtr[t] * kernel[t];
                    ++t;
                }
                t += x;                             // jump to next row
            }

            // store output
            *outPtr = (unsigned char)((float)fabs(sum) + 0.5f);
            ++outPtr;
            ++x;
            for(k=0; k < kSize; ++k) ++inPtr[k];    // move all cursors to next
        }

        ++y;                                        // add one more row to convolve for next run
    }

    // convolve rows from index=kCenterY to index=(dataSizeY-kCenterY-1)
    for(i= kCenterY; i < rowEnd; ++i)               // number of rows
    {
        // partition #4 ***********************************
        x = kCenterX;
        for(j=0; j < kCenterX; ++j)                 // column from index=0 to index=kCenterX-1
        {
            sum = 0;
            t = 0;
            for(m=0; m < kernelSizeY; ++m)
            {
                for(n=0; n <= x; ++n)
                {
                    sum += *inPtr[t] * kernel[t];
                    ++t;
                }
                t += (kernelSizeX - x - 1);
            }

            // store output
            *outPtr = (unsigned char)((float)fabs(sum) + 0.5f);
            ++outPtr;
            ++x;
            for(k=0; k < kSize; ++k) ++inPtr[k];    // move all cursors to next
        }

        // partition #5 ***********************************
        for(j = kCenterX; j < colEnd; ++j)          // column from index=kCenterX to index=(dataSizeX-kCenterX-1)
        {
            sum = 0;
            t = 0;
            for(m=0; m < kernelSizeY; ++m)
            {
                for(n=0; n < kernelSizeX; ++n)
                {
                    sum += *inPtr[t] * kernel[t];
                    ++inPtr[t]; // in this partition, all cursors are used to convolve. moving cursors to next is safe here
                    ++t;
                }
            }

            // store output
            *outPtr = (unsigned char)((float)fabs(sum) + 0.5f);
            ++outPtr;
            ++x;
        }

        // partition #6 ***********************************
        x = 1;
        for(j=colEnd; j < dataSizeX; ++j)           // column from index=(dataSizeX-kCenter) to index=(dataSizeX-1)
        {
            sum = 0;
            t = x;
            for(m=0; m < kernelSizeY; ++m)
            {
                for(n=x; n < kernelSizeX; ++n)
                {
                    sum += *inPtr[t] * kernel[t];
                    ++t;
                }
                t += x;
            }

            // store output
            *outPtr = (unsigned char)((float)fabs(sum) + 0.5f);
            ++outPtr;
            ++x;
            for(k=0; k < kSize; ++k) ++inPtr[k];    // move all cursors to next
        }
    }

    // convolve rows from index=(dataSizeY-kCenterY) to index=(dataSizeY-1)
    y = 1;
    for(i= rowEnd; i < dataSizeY; ++i)               // number of rows
    {
        // partition #7 ***********************************
        x = kCenterX;
        for(j=0; j < kCenterX; ++j)                 // column from index=0 to index=kCenterX-1
        {
            sum = 0;
            t = kernelSizeX * y;

            for(m=y; m < kernelSizeY; ++m)
            {
                for(n=0; n <= x; ++n)
                {
                    sum += *inPtr[t] * kernel[t];
                    ++t;
                }
                t += (kernelSizeX - x - 1);
            }

            // store output
            *outPtr = (unsigned char)((float)fabs(sum) + 0.5f);
            ++outPtr;
            ++x;
            for(k=0; k < kSize; ++k) ++inPtr[k];    // move all cursors to next
        }

        // partition #8 ***********************************
        for(j=kCenterX; j < colEnd; ++j)            // column from index=kCenterX to index=(dataSizeX-kCenterX-1)
        {
            sum = 0;
            t = kernelSizeX * y;
            for(m=y; m < kernelSizeY; ++m)
            {
                for(n=0; n < kernelSizeX; ++n)
                {