Convert HImage to byte[] as File.ReadAllBytes() does with .png

HalconDotNet.HOperatorSet.ReadImage(out HObject image, srcPath);
//...
//(graphic stuff)
//...
HalconDotNet.HOperatorSet.WriteImage(imagePart, "png", 0, tmpImgPath); // skip these steps
Image = File.ReadAllBytes(path)                                        // skip these steps

This piece of code is executed thousands of times. The last two steps are just there to have a compatibility step in between Halcon and .NET as I dont know how to combine them.

What I need is a way to convert a HImage(HObject) to a byte[], the same way WriteImage() + File.ReadAllBytes(path) would do. This last bit is important as this piece of code generates inputs for image classification models.

As the models are trained with data loaded from disk with File.ReadAllBytes(path) I'm assuming I need to prepare the data in the same way when using the model. When I read a 100x100 color PNG (solid color) with File.ReadAllBytes() I don't get 100x100x3 bytes, but 342, so I'm assuming the data is still compressed, and further assuming that I need to guarantee similar data when using the model.

This question has some overlap with this one but I need a byte[] instead of bitmap and just can't get it to work.

an other way could be to convert your hImage to a bitmap and then serialize this bitmap as a byte array.

heres a function to convert a hImage to a System.Drawing.Bitmap

public static Bitmap HImageToBitmap(HImage ho_Image) {
int iWidth, iHeight, iNumChannels;
IntPtr ip_R, ip_G, ip_B, ip_Gray;
String sType;
// null return object
Bitmap bitmap = null;
try {
//
// Note that pixel data is stored differently in System.Drawing.Bitmap
// a) Stride:
// stride is the width, rounded up to a multiple of 4 (padding)
// Size of data array HALCON: heigth*width, Bitmap: heigth*stride
// compare: https://msdn.microsoft.com/en-us/library/zy1a2d14%28v=vs.110%29.aspx
// b) RGB data
// HALCON: three arrays, Bitmap: one array (alternating red/green/blue)
iNumChannels = ho_Image.CountChannels();
if (iNumChannels != 1 && iNumChannels != 3)
throw new Exception("Conversion of HImage to Bitmap failed. Number of channels of the HImage is: " +
iNumChannels + ". Conversion rule exists only for images with 1 or 3 chanels");
if (iNumChannels == 1) {
//
// 1) Get the image pointer
ip_Gray = ho_Image.GetImagePointer1(out sType, out iWidth, out iHeight);
//
// 2) Calculate the stride
int iPadding = (4 - (iWidth % 4)) % 4;
int iStride = iWidth + iPadding;
//
// 3) Create a new gray Bitmap object, allocating the necessary (managed) memory 
bitmap = new Bitmap(iWidth, iHeight, PixelFormat.Format8bppIndexed);
// note for high performance: in case of padding=0, image can be copied by reference.
// however, then the bitmap's validity relies on the HImage lifetime.
// bitmap = new Bitmap(iWidth, iHeight, iWidth, PixelFormat.Format8bppIndexed, ip_Gray);
//
// 4) Copy the image data directly into the bitmap data object, re-arranged in the required bitmap order
// BitmapData lets us access the data in memory
BitmapData bmpData = bitmap.LockBits(new Rectangle(0, 0, iWidth, iHeight),
ImageLockMode.WriteOnly, bitmap.PixelFormat);
// System.Threading.Tasks.Parallel processing requires .NET framework >= 4.0 
Parallel.For(0, iHeight, r => {
IntPtr posRead = ip_Gray + r * iWidth;
IntPtr posWrite = bmpData.Scan0 + r * iStride;
for (int c = 0; c < iWidth; c++)
Marshal.WriteByte((IntPtr)posWrite, c, Marshal.ReadByte((IntPtr)posRead, c));
});
//
// 5) Let the windows memory management take over control
bitmap.UnlockBits(bmpData);
//
// 6) Adjust palette to grayscale (linearized grayscale)
// ColorPalette has no constructor -> obtain it from the static member
ColorPalette cp_P = bitmap.Palette;
for (int i = 0; i < 256; i++) {
cp_P.Entries[i] = Color.FromArgb(i, i, i);
}
bitmap.Palette = cp_P;
}
if (iNumChannels == 3) {
//
// 1) Get the image pointer
ho_Image.GetImagePointer3(out ip_R, out ip_G, out ip_B, out sType, out iWidth, out iHeight);
//
// 2) Calculate the stride
int iPadding = (4 - ((iWidth * 3) % 4)) % 4;
int iStride = iWidth * 3 + iPadding;
//
// 3) Create a new RGB Bitmap object, allocating the necessary (managed) memory 
bitmap = new Bitmap(iWidth, iHeight, PixelFormat.Format24bppRgb);
//
// 4) Copy the image data directly into the bitmap data object, re-arranged in the required bitmap order
// BitmapData lets us access the data in memory
BitmapData bmpData = bitmap.LockBits(new Rectangle(0, 0, iWidth, iHeight),
ImageLockMode.WriteOnly, bitmap.PixelFormat);
Parallel.For(0, iHeight, r => {
IntPtr posReadR = (IntPtr)((long)ip_R + r * iWidth);
IntPtr posReadG = (IntPtr)((long)ip_G + r * iWidth);
IntPtr posReadB = (IntPtr)((long)ip_B + r * iWidth);
IntPtr posWrite = (IntPtr)((long)bmpData.Scan0 + r * iStride);
for (int c = 0; c < iWidth; c++) {
Marshal.WriteByte(posWrite, 3 * c, Marshal.ReadByte(posReadB, c));
Marshal.WriteByte(posWrite, 3 * c + 1, Marshal.ReadByte(posReadG, c));
Marshal.WriteByte(posWrite, 3 * c + 2, Marshal.ReadByte(posReadR, c));
}
});
//
// 5) Let the windows memory management take over control
bitmap.UnlockBits(bmpData);
}
}
catch (Exception ex) {
throw new Exception("Conversion of HImage to Bitmap failed.", ex);
}
return bitmap;
}

Can you try?

public static byte[] ImageToByte(Image image)
{
using (var stream = new MemoryStream())
{
image.Save(stream, System.Drawing.Imaging.ImageFormat.Png);
return stream.ToArray();
}
}

Or

Image image = Image.FromFile(imagePath);
byte[] byteArr = new byte[] { };
using (var ms = new MemoryStream())
{
image.Save(ms, image.RawFormat);
byteArr = ms.ToArray();
}

For Image you will need System.Drawing