views:

766

answers:

2

I'm using Ogre3D and PhysX.

When I load a terrain from an 8bit height map, it looks normal on Visual Debugger.

Look at first image: http://img44.imageshack.us/gal.php?g=44927650.jpg

But when I save the height map as a 16bit image, I get what you see on second image.

Here's the code, thats works normal with 8bit PNG:

    mSceneMgr->setWorldGeometry("terrain.cfg" );
    mTSM=static_cast<TerrainSceneManager*>(sceneMgr);

    TerrainOptions mTerrainOptions = mTSM->getOptions();

    //load heihgtmap
    Image mImage;
    mImage.load("isl_h_ph.png", ResourceGroupManager::getSingleton().getWorldResourceGroupName()); //load image

    //write image buffer to pOrigSrc 
    const uchar* pOrigSrc = mImage.getData();
    const uchar* pSrc;
    // image size to mPageSize
    size_t   mPageSize  = mTerrainOptions.pageSize;

    NxActorDesc ActorDesc;

    //set number of segments
    heightFieldDesc = new NxHeightFieldDesc;
    heightFieldDesc->nbColumns      = mPageSize;
    heightFieldDesc->nbRows         = mPageSize;
    heightFieldDesc->verticalExtent   = -1000; 
    heightFieldDesc->convexEdgeThreshold    = 0;

    heightFieldDesc->samples      = new NxU32[mPageSize*mPageSize];   //constructor for every sample?
    heightFieldDesc->sampleStride   = sizeof(NxU32);                    //some sample step = number of samples

    pSrc = pOrigSrc;
    char* currentByte = (char*)heightFieldDesc->samples;     //current sample mb?
    LogManager::getSingletonPtr()->logMessage("+++Heightmap---");
    for (NxU32 row = 0; row < mPageSize; row++)
    {
        for (NxU32 column = 0; column < mPageSize; column++)   //cycle around samples
        {
            pSrc = pOrigSrc + column*mPageSize +row;

            //NxReal s = NxReal(row) / NxReal(mPageSize);
            //NxReal t = NxReal(column) / NxReal(mPageSize);

            NxI16 height = (NxI32)(*pSrc++);
        NxU32 matrixOffset = (row % gMatrixSize) * gMatrixSize + (column % gMatrixSize);

            //LogManager::getSingletonPtr()->logMessage(Ogre::StringConverter::toString(height));
            NxHeightFieldSample* currentSample = (NxHeightFieldSample*)currentByte;
            currentSample->height = height;
        currentSample->materialIndex0 = gMatrix[matrixOffset][1];
        currentSample->materialIndex1 = gMatrix[matrixOffset][2];
        currentSample->tessFlag = gMatrix[matrixOffset][0];
            currentByte += heightFieldDesc->sampleStride;
        }
    }

    heightField = mScene->getPhysicsSDK().createHeightField(*heightFieldDesc);

    NxHeightFieldShapeDesc heightFieldShapeDesc;
    heightFieldShapeDesc.heightField   = heightField;
    heightFieldShapeDesc.shapeFlags      = NX_SF_FEATURE_INDICES | NX_SF_VISUALIZATION;
    heightFieldShapeDesc.group   = 1;
    heightFieldShapeDesc.heightScale   = 18.8f;//1 в Physx = 255 в огре
    heightFieldShapeDesc.rowScale      = mTerrainOptions.scale.x;
    heightFieldShapeDesc.columnScale   = mTerrainOptions.scale.z;
    heightFieldShapeDesc.meshFlags   = NX_MESH_SMOOTH_SPHERE_COLLISIONS;
    heightFieldShapeDesc.materialIndexHighBits = 0;
    heightFieldShapeDesc.holeMaterial = 2;
    ActorDesc.shapes.pushBack(&heightFieldShapeDesc);

What should I change to get 16bit or higher image loaded working?

P.S.: Sorry for bad English

+1  A: 

Your pOrigSrc is a uchar, which is 8 bits, so you're not getting the correct offset when you do this:

pSrc = pOrigSrc + column*mPageSize +row;

You can fix this by first grabbing the stride of your image before your loops, something like this:

int imageStride = mImage.getBPP() / 8;

and then multiply your calculated offset by the stride, something like this:

pSrc = pOrigSrc + (column*mPageSize +row)*imageStride;

That should allow you to use both 8-bit and 16-bit height maps. PhysX only supports 16-bit height maps, so you can't go higher than that.

Gerald
thanks now its look a little bit nice =)but still no goodhttp://img34.imageshack.us/i/new161.jpg/
Neonai
A: 

All of the chars in your code need to now be shorts, and you need to traverse the file 2 bytes at a time, not 1. Because 8 bits is one byte (the size of a char or uchar), and 16 bits is two bytes, the size of a short or unsigned short.

Ricket
how to get short if getData return only uchar
Neonai