示例#1
0
void faster_corner_detect_10(const IplImage* I, mrpt::vision::TSimpleFeatureList & corners, int barrier, uint8_t octave, std::vector<size_t> * out_feats_index_by_row)
{
	MRPT_UNUSED_PARAM(octave);
	corners.reserve(corners.size()+500);
	//corners.mark_kdtree_as_outdated();

	size_t *ptr_feat_index_by_row;
	if (out_feats_index_by_row)
	{
		out_feats_index_by_row->resize(I->height);
		ptr_feat_index_by_row = &(*out_feats_index_by_row)[0];
	}
	else {
		ptr_feat_index_by_row = NULL;
	}


	// 3 first rows have no features:
	if (ptr_feat_index_by_row) {
		*ptr_feat_index_by_row++ = corners.size();
		*ptr_feat_index_by_row++ = corners.size();
		*ptr_feat_index_by_row++ = corners.size();
	}

	const int w = I->width;
	const int stride = 3*I->widthStep; // 3*w;

	// The compiler refuses to reserve a register for this
	register const __m128i barriers = _mm_set1_epi8((uint8_t)barrier);

	int xend = I->width - 3;
	xend -= (I->width-3) % 16;

	for(int y=3; y < I->height - 3; y++)
	{
		if (ptr_feat_index_by_row)  // save index by row:
			*ptr_feat_index_by_row++=corners.size();

		for(int x=3; x < 16; x++)
	    	if(is_corner_10<Less>((const uint8_t*)I->imageData+I->widthStep*y+x, I->widthStep, barrier) || is_corner_10<Greater>((const uint8_t*)I->imageData+I->widthStep*y+x, I->widthStep, barrier))
		    corners.push_back_fast(x, y);

	    for(int x=16; x < xend; x+=16)
	    {
	    	const uint8_t* p = (const uint8_t*)I->imageData+I->widthStep*y+x;
		__m128i lo, hi;
		{
		    const __m128i here = load_si128<Aligned>((const __m128i*)(p));
		    lo = _mm_subs_epu8(here, barriers);
		    hi = _mm_adds_epu8(barriers, here);
		}
		unsigned int ans_b, ans_e;
		{
		    __m128i top = load_si128<Aligned>((const __m128i*)(p-stride));
		    __m128i bottom = load_si128<Aligned>((const __m128i*)(p+stride));

		    CHECK_BARRIER(lo, hi, top, ans_b);
		    CHECK_BARRIER(lo, hi, bottom, ans_e);
		    if (!(ans_b | ans_e))
			continue;
		}

		unsigned int ans_m, ans_p, possible;
		{
		    __m128i ul = _mm_loadu_si128((const __m128i*)(p-2-2*w));
		    __m128i lr = _mm_loadu_si128((const __m128i*)(p+2+2*w));
		    CHECK_BARRIER(lo, hi, ul, ans_m);
		    CHECK_BARRIER(lo, hi, lr, ans_p);
		    possible = (ans_m & ans_b) | (ans_e & ans_p);
		    if (!possible)
			continue;
		}

		unsigned int ans_o, ans_n;
		{
		    __m128i ll = _mm_loadu_si128((const __m128i*)(p-2+2*w));
		    __m128i ur = _mm_loadu_si128((const __m128i*)(p+2-2*w));
		    CHECK_BARRIER(lo, hi, ll, ans_o);
		    CHECK_BARRIER(lo, hi, ur, ans_n);
		    possible &= ans_o | (ans_b & ans_n);
		    possible &= ans_n | (ans_e & ans_o);
		    if (!possible)
			continue;
		}

		unsigned int ans_h, ans_k;
		{
		    __m128i left = _mm_loadu_si128((const __m128i*)(p-3));
		    __m128i right = _mm_loadu_si128((const __m128i*)(p+3));
		    CHECK_BARRIER(lo, hi, left, ans_h);
		    CHECK_BARRIER(lo, hi, right, ans_k);
		    possible &= ans_h | (ans_n & ans_k & ans_p);
		    possible &= ans_k | (ans_m & ans_h & ans_o);
		    if (!possible)
			continue;
		}

		unsigned int ans_a, ans_c;
		{
		    __m128i a = _mm_loadu_si128((const __m128i*)(p-1-stride));
		    __m128i c = _mm_insert_epi16(_mm_srli_si128(a,2), *(const unsigned short*)(p+15-stride), 7);
		    //__m128i c = _mm_loadu_si128((const __m128i*)(p+1-stride));
		    CHECK_BARRIER(lo, hi, a, ans_a);
		    CHECK_BARRIER(lo, hi, c, ans_c);
		    possible &= ans_a | (ans_e & ans_p);
		    possible &= ans_c | (ans_o & ans_e);
		    if (!possible)
			continue;
		}

		unsigned int ans_d, ans_f;
		{
		    __m128i d = _mm_loadu_si128((const __m128i*)(p-1+stride));
		    __m128i f = _mm_insert_epi16(_mm_srli_si128(d,2), *(const unsigned short*)(p+15+stride), 7);
		    //__m128i f = _mm_loadu_si128((const __m128i*)(p+1+stride));
		    CHECK_BARRIER(lo, hi, d, ans_d);
		    CHECK_BARRIER(lo, hi, f, ans_f);
		    const unsigned int ans_abc = ans_a & ans_b & ans_c;
		    possible &= ans_d | (ans_abc & ans_n);
		    possible &= ans_f | (ans_m & ans_abc);
		    if (!possible)
			continue;
		}

		unsigned int ans_g, ans_i;
		{
		    __m128i g = _mm_loadu_si128((const __m128i*)(p-3-w));
		    __m128i ii = _mm_loadu_si128((const __m128i*)(p-3+w));
		    CHECK_BARRIER(lo, hi, g, ans_g);
		    CHECK_BARRIER(lo, hi, ii, ans_i);
		    possible &= ans_g | (ans_f & ans_p & ans_k);
		    possible &= ans_i | (ans_c & ans_n & ans_k);
		    if (!possible)
			continue;
		}

		unsigned int ans_j, ans_l;
		{
		    __m128i jj = _mm_loadu_si128((const __m128i*)(p+3-w));
		    __m128i l = _mm_loadu_si128((const __m128i*)(p+3+w));
		    CHECK_BARRIER(lo, hi, jj, ans_j);
		    CHECK_BARRIER(lo, hi, l, ans_l);
		    const unsigned int ans_ghi = ans_g & ans_h & ans_i;
		    possible &= ans_j | (ans_d & ans_o & ans_ghi);
		    possible &= ans_l | (ans_m & ans_a & ans_ghi);
		    if (!possible)
			continue;
		}

		possible |= (possible >> 16);
		//if(possible & 0x0f) //Does this make it faster?
		{
		    if(possible & (1<< 0))
		      corners.push_back_fast(x + 0, y);
		    if(possible & (1<< 1))
		      corners.push_back_fast(x + 1, y);
		    if(possible & (1<< 2))
		      corners.push_back_fast(x + 2, y);
		    if(possible & (1<< 3))
		      corners.push_back_fast(x + 3, y);
		    if(possible & (1<< 4))
		      corners.push_back_fast(x + 4, y);
		    if(possible & (1<< 5))
		      corners.push_back_fast(x + 5, y);
		    if(possible & (1<< 6))
		      corners.push_back_fast(x + 6, y);
		    if(possible & (1<< 7))
		      corners.push_back_fast(x + 7, y);
		}
		//if(possible & 0xf0) //Does this mak( ,  fast)r?
		{
		    if(possible & (1<< 8))
		      corners.push_back_fast(x + 8, y);
		    if(possible & (1<< 9))
		      corners.push_back_fast(x + 9, y);
		    if(possible & (1<<10))
		      corners.push_back_fast(x +10, y);
		    if(possible & (1<<11))
		      corners.push_back_fast(x +11, y);
		    if(possible & (1<<12))
		      corners.push_back_fast(x +12, y);
		    if(possible & (1<<13))
		      corners.push_back_fast(x +13, y);
		    if(possible & (1<<14))
		      corners.push_back_fast(x +14, y);
		    if(possible & (1<<15))
		      corners.push_back_fast(x +15, y);
		}
	    }

	    for(int x=xend; x < I->width - 3; x++)
	    	if(is_corner_10<Less>((const uint8_t*)I->imageData+I->widthStep*y+x, I->widthStep, barrier) || is_corner_10<Greater>((const uint8_t*)I->imageData+I->widthStep*y+x, I->widthStep, barrier))
		    corners.push_back_fast(x, y);
	}

	// 3 last rows have no features:
	if (ptr_feat_index_by_row) {
		*ptr_feat_index_by_row++ = corners.size();
		*ptr_feat_index_by_row++ = corners.size();
		*ptr_feat_index_by_row++ = corners.size();
	}

}
示例#2
0
void faster_corner_detect_9(const IplImage* I, mrpt::vision::TSimpleFeatureList & corners, int barrier, uint8_t octave, std::vector<size_t> * out_feats_index_by_row)
{
	corners.reserve(corners.size()+500);
	//corners.mark_kdtree_as_outdated();

	size_t *ptr_feat_index_by_row;
	if (out_feats_index_by_row)
	{
		out_feats_index_by_row->resize(I->height);
		ptr_feat_index_by_row = &(*out_feats_index_by_row)[0];
	}
	else {
		ptr_feat_index_by_row = NULL;
	}

	const int w = I->width;
	const int stride = 3*I->widthStep; // 3*w;

	// The compiler refuses to reserve a register for this
	register const __m128i barriers = _mm_set1_epi8((uint8_t)barrier);

	int xend = I->width - 3;
	xend -= (I->width-3) % 16;

	// 3 first rows have no features:
	if (ptr_feat_index_by_row) {
		*ptr_feat_index_by_row++ = corners.size();
		*ptr_feat_index_by_row++ = corners.size();
		*ptr_feat_index_by_row++ = corners.size();
	}

	for(int y=3; y < I->height - 3; y++)
	{
		if (ptr_feat_index_by_row)  // save index by row:
			*ptr_feat_index_by_row++=corners.size();

		for(int x=3; x < 16; x++)
			if(is_corner_9<Less>( (const uint8_t*)I->imageData+I->widthStep*y+x, I->widthStep, barrier) || is_corner_9<Greater>((const uint8_t*)I->imageData+I->widthStep*y+x, I->widthStep, barrier))
		    corners.push_back_fast(x<<octave, y<<octave);

	    for(int x=16; x < xend; x+=16)
	    {
	    	const uint8_t* p = (const uint8_t*)I->imageData+I->widthStep*y+x; //(const uint8_t*)I->imageData+I->widthStep*y+x;
		__m128i lo, hi;
		{
		    const __m128i here = load_si128<Aligned>((const __m128i*)(p));
		    lo = _mm_subs_epu8(here, barriers);
		    hi = _mm_adds_epu8(barriers, here);
		}
		unsigned int ans_0, ans_8, possible;
		{
		    __m128i top = load_si128<Aligned>((const __m128i*)(p-stride));
		    __m128i bottom = load_si128<Aligned>((const __m128i*)(p+stride));

		    CHECK_BARRIER(lo, hi, top, ans_0);
		    CHECK_BARRIER(lo, hi, bottom, ans_8);
		    possible = ans_0 | ans_8;
		    if (!possible)
			continue;
		}

		unsigned int ans_15, ans_1;
		{
		    __m128i a = _mm_loadu_si128((const __m128i*)(p-1-stride));
		    __m128i c = _mm_insert_epi16(_mm_srli_si128(a,2), *(const unsigned short*)(p+15-stride), 7);
		    CHECK_BARRIER(lo, hi, a, ans_15);
		    CHECK_BARRIER(lo, hi, c, ans_1);
		    possible &= ans_8 | (ans_15 & ans_1);
		    if (!possible)
			continue;
		}

		unsigned int ans_9, ans_7;
		{
		    __m128i d = _mm_loadu_si128((const __m128i*)(p-1+stride));
		    __m128i f = _mm_insert_epi16(_mm_srli_si128(d,2), *(const unsigned short*)(p+15+stride), 7);
		    CHECK_BARRIER(lo, hi, d, ans_9);
		    CHECK_BARRIER(lo, hi, f, ans_7);
		    possible &= ans_9 | (ans_0 & ans_1);
		    possible &= ans_7 | (ans_15 & ans_0);
		    if (!possible)
			continue;
		}

		unsigned int ans_12, ans_4;
		{
		    __m128i left = _mm_loadu_si128((const __m128i*)(p-3));
		    __m128i right = _mm_loadu_si128((const __m128i*)(p+3));
		    CHECK_BARRIER(lo, hi, left, ans_12);
		    CHECK_BARRIER(lo, hi, right, ans_4);
		    possible &= ans_12 | (ans_4 & (ans_1 | ans_7));
		    possible &= ans_4 | (ans_12 & (ans_9 | ans_15));
		    if (!possible)
			continue;
		}

		unsigned int ans_14, ans_6;
		{
		    __m128i ul = _mm_loadu_si128((const __m128i*)(p-2-2*w));
		    __m128i lr = _mm_loadu_si128((const __m128i*)(p+2+2*w));
		    CHECK_BARRIER(lo, hi, ul, ans_14);
		    CHECK_BARRIER(lo, hi, lr, ans_6);
		    {
			const unsigned int ans_6_7 = ans_6 & ans_7;
			possible &= ans_14 | (ans_6_7 & (ans_4 | (ans_8 & ans_9)));
			possible &= ans_1 | (ans_6_7) | ans_12;
		    }
		    {
			const unsigned int ans_14_15 = ans_14 & ans_15;
			possible &= ans_6 | (ans_14_15 & (ans_12 | (ans_0 & ans_1)));
			possible &= ans_9 | (ans_14_15) | ans_4;
		    }
		    if (!possible)
			continue;
		}

		unsigned int ans_10, ans_2;
		{
		    __m128i ll = _mm_loadu_si128((const __m128i*)(p-2+2*w));
		    __m128i ur = _mm_loadu_si128((const __m128i*)(p+2-2*w));
		    CHECK_BARRIER(lo, hi, ll, ans_10);
		    CHECK_BARRIER(lo, hi, ur, ans_2);
		    {
			const unsigned int ans_1_2 = ans_1 & ans_2;
			possible &= ans_10 | (ans_1_2 & ((ans_0 & ans_15) | ans_4));
			possible &= ans_12 | (ans_1_2) | (ans_6 & ans_7);
		    }
		    {
			const unsigned int ans_9_10 = ans_9 & ans_10;
			possible &= ans_2 | (ans_9_10 & ((ans_7 & ans_8) | ans_12));
			possible &= ans_4 | (ans_9_10) | (ans_14 & ans_15);
		    }
		    possible &= ans_8 | ans_14 | ans_2;
		    possible &= ans_0 | ans_10 | ans_6;
		    if (!possible)
			continue;
		}

		unsigned int ans_13, ans_5;
		{
		    __m128i g = _mm_loadu_si128((const __m128i*)(p-3-w));
		    __m128i l = _mm_loadu_si128((const __m128i*)(p+3+w));
		    CHECK_BARRIER(lo, hi, g, ans_13);
		    CHECK_BARRIER(lo, hi, l, ans_5);
		    const unsigned int ans_15_0 = ans_15 & ans_0;
		    const unsigned int ans_7_8 = ans_7 & ans_8;
		    {
			const unsigned int ans_12_13 = ans_12 & ans_13;
			possible &= ans_5 | (ans_12_13 & ans_14 & ((ans_15_0) | ans_10));
			possible &= ans_7 | (ans_1 & ans_2) | (ans_12_13);
			possible &= ans_2 | (ans_12_13) | (ans_7_8);
		    }
		    {
			const unsigned int ans_4_5 = ans_4 & ans_5;
			const unsigned int ans_9_10 = ans_9 & ans_10;
			possible &= ans_13 | (ans_4_5 & ans_6 & ((ans_7_8) | ans_2));
			possible &= ans_15 | (ans_4_5) | (ans_9_10);
			possible &= ans_10 | (ans_4_5) | (ans_15_0);
			possible &= ans_15 | (ans_9_10) | (ans_4_5);
		    }

		    possible &= ans_8 | (ans_13 & ans_14) | ans_2;
		    possible &= ans_0 | (ans_5 & ans_6) | ans_10;
		    if (!possible)
			continue;
		}


		unsigned int ans_11, ans_3;
		{
		    __m128i ii = _mm_loadu_si128((const __m128i*)(p-3+w));
		    __m128i jj = _mm_loadu_si128((const __m128i*)(p+3-w));
		    CHECK_BARRIER(lo, hi, ii, ans_11);
		    CHECK_BARRIER(lo, hi, jj, ans_3);
		    {
			const unsigned int ans_2_3 = ans_2 & ans_3;
			possible &= ans_11 | (ans_2_3 & ans_4 & ((ans_0 & ans_1) | (ans_5 & ans_6)));
			possible &= ans_13 | (ans_7 & ans_8) | (ans_2_3);
			possible &= ans_8 | (ans_2_3) | (ans_13 & ans_14);
		    }
		    {
			const unsigned int ans_11_12 = ans_11 & ans_12;
			possible &= ans_3 | (ans_10 & ans_11_12 & ((ans_8 & ans_9) | (ans_13 & ans_14)));
			possible &= ans_1 | (ans_11_12) | (ans_6 & ans_7);
			possible &= ans_6 | (ans_0 & ans_1) | (ans_11_12);
		    }
		    {
			const unsigned int ans_3_4 = ans_3 & ans_4;
			possible &= ans_9 | (ans_3_4) | (ans_14 & ans_15);
			possible &= ans_14 | (ans_8 & ans_9) | (ans_3_4);
		    }
		    {
			const unsigned int ans_10_11 = ans_10 & ans_11;
			possible &= ans_5 | (ans_15 & ans_0) | (ans_10_11);
			possible &= ans_0 | (ans_10_11) | (ans_5 & ans_6);
		    }
		    if (!possible)
			continue;

		}

		possible |= (possible >> 16);

		//if(possible & 0x0f) //Does this make it faster?
		{
		    if(possible & (1<< 0))
		      corners.push_back_fast((x + 0)<<octave, y<<octave);
		    if(possible & (1<< 1))
		      corners.push_back_fast((x + 1)<<octave, y<<octave);
		    if(possible & (1<< 2))
		      corners.push_back_fast((x + 2)<<octave, y<<octave);
		    if(possible & (1<< 3))
		      corners.push_back_fast((x + 3)<<octave, y<<octave);
		    if(possible & (1<< 4))
		      corners.push_back_fast((x + 4)<<octave, y<<octave);
		    if(possible & (1<< 5))
		      corners.push_back_fast((x + 5)<<octave, y<<octave);
		    if(possible & (1<< 6))
		      corners.push_back_fast((x + 6)<<octave, y<<octave);
		    if(possible & (1<< 7))
		      corners.push_back_fast((x + 7)<<octave, y<<octave);
		}
		//if(possible & 0xf0) //Does this mak( ,  fast)r?
		{
		    if(possible & (1<< 8))
		      corners.push_back_fast((x + 8)<<octave, y<<octave);
		    if(possible & (1<< 9))
		      corners.push_back_fast((x + 9)<<octave, y<<octave);
		    if(possible & (1<<10))
		      corners.push_back_fast((x +10)<<octave, y<<octave);
		    if(possible & (1<<11))
		      corners.push_back_fast((x +11)<<octave, y<<octave);
		    if(possible & (1<<12))
		      corners.push_back_fast((x +12)<<octave, y<<octave);
		    if(possible & (1<<13))
		      corners.push_back_fast((x +13)<<octave, y<<octave);
		    if(possible & (1<<14))
		      corners.push_back_fast((x +14)<<octave, y<<octave);
		    if(possible & (1<<15))
		      corners.push_back_fast((x +15)<<octave, y<<octave);
		}
	    }

	    for(int x=xend; x < I->width - 3; x++)
			if(is_corner_9<Less>((const uint8_t*)I->imageData+I->widthStep*y+x, I->widthStep, barrier) || is_corner_9<Greater>((const uint8_t*)I->imageData+I->widthStep*y+x, I->widthStep, barrier))
		    corners.push_back_fast(x<<octave, y<<octave);
	}

	// 3 last rows have no features:
	if (ptr_feat_index_by_row) {
		*ptr_feat_index_by_row++ = corners.size();
		*ptr_feat_index_by_row++ = corners.size();
		*ptr_feat_index_by_row++ = corners.size();
	}

}