Cast-Aways

Make your constants double at compile time.

BubbleFrame[b, a, c] = 0D;, 0.12D, BubbleFrame[b + 1, a - 1, c - 1] = 0D etc. Well, 0.12 is already a double, but what the heck.

BubbleCounter is used as a value in the matrix. Declare it as double and pass it as a double.

Guessing At Performance

Not this:

 calibline= BubbleReader.ReadLine();
 numbers = calibline.Split(' ');
 for (int a = 0; a < Sensor; a++)
 {
    BubbleFrame[b, a, c] = Convert.ToDouble(numbers[a]);
 }

but this:

BubbleFrame[b,a] = Array.ConvertAll(BubbleReader.ReadLine().Split(' '), double.Parse);

I honestly don't know that if this is more performant. And I hope the new array does not have > Sensor values!

Why this?

Maxvalue = BubbleFrame.Cast<double>().Max();

Aren't BubbleFrame values already double?

Calculate once; and avoid unnecessary call:

Cminus1 = c-1;

if (Cminus1 > 0)
    BubbleSearchCMinus1(ref BubbleFrame, ref BubbleCollection, ref Cminus1, ref b, ref a, ref Threshold, ref BubbleCounter, ref Sensor, ref n_fr);

then inside:

 BubbleSearchCMinus1(ref double[, ,] BubbleFrame,ref double[, ,] BubbleCollection, ref int c, ref int b, ref int a, ref double Threshold, ref int BubbleCounter, ref int Sensor, ref int n_fr) {
    Bminus1 = b-1;  Bplus1 = b+1;
}

Cast-Aways

Make your constants double at compile time.

BubbleFrame[b, a, c] = 0D;, 0.12D, BubbleFrame[b + 1, a - 1, c - 1] = 0D etc. Well, 0.12 is already a double, but what the heck.

BubbleCounter is used as a value in the matrix. Declare it as double and pass it as a double.

Guessing At Performance

Not this:

 calibline= BubbleReader.ReadLine();
 numbers = calibline.Split(' ');
 for (int a = 0; a < Sensor; a++)
 {
    BubbleFrame[b, a, c] = Convert.ToDouble(numbers[a]);
 }

but this:

BubbleFrame[b,a] = Array.ConvertAll(BubbleReader.ReadLine().Split(' '), double.Parse);

I honestly don't know that this is more performant. And I hope the new array does not have > Sensor values!

Why this?

Maxvalue = BubbleFrame.Cast<double>().Max();

Aren't BubbleFrame values already double?

Calculate once; and avoid unnecessary call:

Cminus1 = c-1;

if (Cminus1 > 0)
    BubbleSearchCMinus1(ref BubbleFrame, ref BubbleCollection, ref Cminus1, ref b, ref a, ref Threshold, ref BubbleCounter, ref Sensor, ref n_fr);

then inside:

 BubbleSearchCMinus1(ref double[, ,] BubbleFrame,ref double[, ,] BubbleCollection, ref int c, ref int b, ref int a, ref double Threshold, ref int BubbleCounter, ref int Sensor, ref int n_fr) {
    Bminus1 = b-1;  Bplus1 = b+1;
}